SDS 645: Machine Learning for Video Games

Podcast Guest: Carly Taylor

January 17, 2023

Have you ever considered how machine learning affects your gaming experience? If you’re like us, you probably have! That’s why we’re excited to welcome Activision’s Carly Taylor to the SuperDataScience podcast and dive into machine learning for video games. As the Lead ML Engineer for the iconic Call of Duty, Carly uses ML to detect and minimize disruptive behavior in the game, and this week she’s here to shed light on the importance of low-latency, the future of gaming, her thoughts on the metaverse, and so much more.

Thanks to our Sponsors: 
Interested in sponsoring a SuperDataScience Podcast episode? Visit JonKrohn.com/podcast for sponsorship information.
About Carly Taylor
Carly is a data scientist, computational chemist and machine learning enthusiast. She obtained her M.S. in chemistry from the University of Colorado focusing on computational quantum dynamics. She has authored multiple peer-reviewed publications and holds two non-provisional machine learning patents. When she isn’t writing about herself in the third person, building mechanical keyboards or neglecting the oxford comma, she works as a security strategist for Call of Duty at Activision Publishing. 
Overview
Gaming can be all-consuming when you’re in the thick of it, but this week, Carly Taylor delivers a few lightbulb moments that have us thinking differently about the industry and where it’s headed. The self-styled ‘Rebel Data Scientist,’ who works on the wildly popular Call of Duty game, oversees and implements machine learning projects as the expert ML engineer within the franchise. She’s also pulling double duty as the senior manager of security strategy– a role that sees her identifying anomalous behavior.
Another problem that she regularly focuses on is low-latency, which plays a critical role in the online gaming experience. As users demand bigger and more robust games, Carly credits online engineers as vital team members who ensure their smooth operation. Whether it’s optimizing code, storing data for quick retrieval, or understanding the fundamental workings of PySpark or SQL, online engineering and networking play a vital role in pulling off machine learning in real-time.
Gaming is only getting bigger and better, and its future is sure to reach even wider audiences, Carly predicts. In the coming years, you’ll see the industry expanding to mobile as phones become more capable, and its appeal widens to more demographics. With gaming eclipsing Hollywood and the music industry already, its potential for growth is endless, and it’s only a matter of time until further adoption and democratization catapult the industry to new heights.
But Carly’s insights extend far beyond gaming. She highly recommends Python’s SHAP package for explainable AI, and in particular for plotting interaction terms. She also emphasizes the importance of being able to communicate your results if you want to become a ‘trusted collaborative partner” who empowers others. Hear more from Carly, including her thoughts on the Metaverse, her favorite software packages and how to file a patent, by tuning in this episode today. 
In this episode you will learn: 
  • The relationship between data science and cyber security [04:49]
  • The importance of low-latency for an optimal gaming experience [09:15]
  • The future of gaming [18:13]
  • Carly’s thoughts on the Metaverse [25:43]
  • Carly’s favorite operating systems, software packages, and keyboards [30:27]
  • How to transition from a quantitative academic background into data science [45:28]
  • Why Carly’s called the “Rebel Data Scientist” [53:27]
  • How to file a patent [57:21]
 
Items mentioned in this podcast:
Follow Carly:

Podcast Transcript

Jon Krohn: 00:00:00

This is episode number 645 with Carly Taylor, lead machine learning engineer for Activision’s Call of Duty franchise. Today’s episode is brought to you by Kolena, the testing platform for machine learning.
00:00:17
Welcome to the SuperDataScience podcast, the most listened-to podcast in the data science industry. Each week we bring you inspiring people and ideas to help you build a successful career in data science. I’m your host, Jon Krohn. Thanks for joining me today. Now, let’s make the complex simple. 
 
00:00:47
Welcome back to the SuperDataScience Podcast. Today, we’ve got the absolutely fabulous self-styled rebel data scientist, Carly Taylor joining me on the program. Carly grew rapidly from a senior data scientist role to simultaneously holding Expert Machine Learning Engineer and Senior Manager of Security Strategy titles, since joining Activision two years ago. At Activision, she specifically works on Call of Duty, one of the top grossing video game franchises of all time, with over $30 billion in sales and 250 million global users annually. Prior to Activision, Carly rapidly grew from analyst to data scientist roles. She’s amassed a LinkedIn following of 75,000 plus by regularly posting fruitful tips on breaking into a data science career and progressing within it. She advocates for women in STEM, tech, and gaming careers. She offers one-to-one career consulting to anyone who desires it, we’ve got a link in the show notes for that. She holds a master’s in computational chemistry from the University of Colorado, and she also completed the galvanized data science immersive program. 
 
00:01:57
Today’s episode certainly has technical tidbits throughout that will be useful to hands-on practitioners, but much of the wide-ranging conversation will be fascinating to any listener, particularly if you have an interest in video games, the so-called metaverse or real-time machine learning. In this episode, Carly details what the future of gaming holds, she talks about why low-latency is critical for an optimal gaming experience, and the tools that online engineers use to make that online low-latency experience happen. She talks about her favorite operating systems, software packages and keyboards, how to transition effectively from a quantitative academic background into data science, how to file a patent, and why she’s called the “Rebel Data Scientist”. All right, you ready for this rebellious episode? Let’s go. 
 
00:02:52
Yeah. Carly Taylor, welcome to the SuperDataScience Podcast. Where in the world are you calling in from? 
Carly Taylor: 00:02:57
I am calling in from beautiful Santa Monica, California. 
Jon Krohn: 00:03:01
Nice, and so we know each other through a few different people. We’d never had a conversation before, but you’d been showing up on my LinkedIn feed over and over and I was like, “Who is this cool machine learning expert, Carly Taylor, with all these followers, and how come she hasn’t been on the show yet?” Then I finally got the courage together to ask you to be on the program. 
Carly Taylor: 00:03:23
I’m so glad that you did. 
Jon Krohn: 00:03:28
That very same week that I asked you, and you said yes right away. Serg Masís, who is our researcher on the program, I asked him, he’s been researching on the show for at least six months, and I don’t know how, it had never occurred to me to ask him this question before, but I was like, “Serg, is there anyone out there that you think we should definitely have on the show?” And he sent back three names, and one of them was Carly Taylor. 
Carly Taylor: 00:03:56
Serg is such good people. I love that. 
Jon Krohn: 00:04:00
Yeah. Then we found out in chatting just before we started recording that you also know several other guests that have been on the show, like Christina Stathopoulos, who’s in episode number 603, Joe Reis, who is in episode number 595. Oh, and I should mention that Serg has an amazing episode on explainable AI number 539. It’s one not to miss. So, yeah, so lots of connections. Carly, you’re a data scientist at Activision. 
Carly Taylor: 00:04:31
I am. 
Jon Krohn: 00:04:33
So, you work particularly on a video game that a lot of people love. 
Carly Taylor: 00:04:41
One of the best in the world. My unbiased opinion. 
Jon Krohn: 00:04:47
Yeah. So, you work on implementing and overseeing machine learning projects for Call of Duty to detect and minimize disruptive behavior in the game, and you’re also a senior manager of security strategy. So, what do those roles entail? How do they relate to each other? What is the impact for all of those hardcore Call of Duty fans out there? 
Carly Taylor: 00:05:14
Yeah, I’m basically playing like Pokemon with my titles, trying to collect them all. I’m just, “How many can I have?” 
Jon Krohn: 00:05:21
Simultaneously? 
Carly Taylor: 00:05:23
Exactly. Just do all the things. Yep, you hit the nail on the head. I have two titles. I’m an individual contributor, data scientist at Activision, and I’m also a senior manager for a security strategy. Both of those titles really come down to one thing, I am trying to find the best ways to use machine learning to make our game safer and more fun for our players. 
Jon Krohn: 00:05:44
Nice. So, obviously you can’t dig into the particulars of security strategy at your specific firm or you’d be opening up opportunities for security lapses, but you’re working at the intersection of cybersecurity and data science. So, what kinds of problems would a data scientist working in that space be tackling in general? 
Carly Taylor: 00:06:10
Yeah, that’s a great question. This is such a hot space that’s like cybersecurity is growing, right? We’ve had a lot of high-level hacks and things going on in the world that people are becoming more aware of how important it is to have a good cybersecurity strategy, and then we also have data science that continues to just be like, in my opinion, the hottest job ever, and so trying to marry these two things together is a place that I think is really exciting for people to learn and grow. 
 
00:06:39
What I would say is most of what I’ve seen in the data science cyberspace are people working on things like credit card fraud. You have all of these banks who are trying to make sure that when you go to buy popcorn at the movies, your credit card doesn’t get declined. But, if someone goes to buy popcorn at the movies a thousand miles away from where you actually live, they might say that this is a suspicious charge and try to decline the charge or just not let it go through. 
 
00:07:06
There are a ton of people working in the space to try to find anomalous behavior. I remember when I was in my bootcamp and when I was in school, I would always try to drop the outliers from a dataset. That was kind of what you were taught to either if something’s not available just infer zero or drop it completely, and I think honestly, now that I work in this space, some of the most interesting data science problems actually exist within the outliers. Trying to understand outlier behavior and anomalies, why you see anomalies in your dataset, and if they’re true anomalies, how you can flag them and try to get to the root of the cause. 
Jon Krohn: 00:07:47
That is interesting. I had not thought about that. I am still, all the kinds of problems that I’m solving, the outliers are something that I wanted [inaudible 00:07:57], and I hadn’t thought about that, how it’s those kinds of rare anomalies that are actually the most meaningful thing to you potentially. I also liked how you said that data science is the hottest job ever, because in the very next episode of the show, assuming that everything goes to plan with production and recording and everything, episode number 647 will have Tom Davenport, who is the person who coined, 10 years ago, the idea of data science being the sexiest job of the 21st century. 
Carly Taylor: 00:08:26
Wow. A true hero. What a pioneer.
Jon Krohn: 00:08:31
But, yeah. So, yeah, super interesting example there. Yeah, the relationship between cybersecurity and data science, and I also like the point that you made there that these are two very popular fields. So, data science enormously fast-growing field still today, cybersecurity as well, and so if you’re a listener out there who’s thinking of getting started in data science and you’re not sure where to start, you want somewhere that- 
Carly Taylor: 00:09:01
Come on in. 
Jon Krohn: 00:09:01
… is, yeah, doubly hot. 
Carly Taylor: 00:09:03
There’s dozens of us. 
Jon Krohn: 00:09:06
Nice. Yeah, there you go. So, that’s a great tip. All right, so in addition to security, another kind of problem that must be very important in gaming is having things be low-latency. People are playing games online with other people all over the world and milliseconds matter like when somebody’s sniping you in Call of Duty, or you see that grenade flying to the room or whatever, you need to be able to react and have those reactions computed instantaneously and spread around the world. So, yeah, are you able to elaborate a little bit on that and tell us what kinds of roles enable this kind of super low-latency, optimal gaming experience? 
Carly Taylor: 00:09:54
For sure. Yeah, this is such an interesting area for me because we hire so many online engineers and I’ve been looking at the space recently and thinking it used to be that networking, online engineering, network protocols, was a cool space people wanted to go into, and then I think in the last probably 10 years, that’s shifted a little bit because it’s no longer in vogue, people felt like all the problems that needed to get solved with the internet have been solved, for the most part, we understand how to handle networking protocols and everything here is good to go. 
 
00:10:30
I think now with the renewed importance for online games that rely on the backbone of the internet in order to be able to perform like you just said, and what players want and what I want as a player is for games to just keep getting bigger and bigger. To me, the most fun I have is knowing I’m in a lobby with 150 other real people from all over the world and we’re experiencing something together, and I think that’s really incredible. But, the only thing that makes that possible are the people who are working on our online engineering and online networking, and it is such a hot field now that’s going to keep getting bigger. 
 
00:11:08
So, if this is an area that you’re interested in getting into and you’re excited to figure out what makes the games you play possible, I’d say that this is an awesome place to start spending your energy, and the same thing for data science. I thought I knew how to write optimized code, specifically SQL. Like I thought I could write something that would run pretty quickly, and let me tell you, the mistakes that I have made in just not understanding exactly how SQL evaluates, exactly how PySpark is going to work behind the scenes to make something even faster, exactly how to store my data so that you can retrieve it really quickly. I have learned a lot in the pursuit of getting things done quickly. 
Jon Krohn: 00:11:57
That is super interesting, and you have opened my mind just now to a whole new career title. I was not aware of the title Online Engineer. 
Carly Taylor: 00:12:08
Yeah. Yes. We specifically hire online engineers. I work with a lot of them every single day. They’re my favorite people. If you think you have the chops, always looking. 
Jon Krohn: 00:12:20
Cool. There you go. Are you unit testing your machine learning models? You certainly should be. If you’re not, you should check out Kolena. Kolena is an ML testing platform for your computer vision models. It’s the only tool that allows you to run unit and regression tests at the subclass level on your model after every single model update, allowing you to understand the failure modes of your model much faster. That’s not all. Kolena also automates and standardizes your model’s testing workflows, saving over 40% of your team’s valuable time. Head over to Kolena’s website now to learn more. It’s www.kolena.io. That’s K-O-L-E-N-A .io. 
00:13:06
Another thing that could potentially be interesting real-time in a video game, so, in addition to obviously having to have everything be super low-latency for humans to be interacting with each other, are there elements where we need to have machine learning happening in real-time? 
Carly Taylor: 00:13:21
Yes, a hundred percent. I think that there were some big companies that really pioneered this, trying to get us to the forefront of how you would actually do machine learning in real-time, and have been pushing the boundaries luckily for people like me who came after, trying to use things like Kafka streaming and figuring out what exactly does this look like. Yes, I would say that the more that we need to focus on making decisions as quickly as possible, whether that looks like recommending an advertisement to someone or making a decision in a game about something that needs to happen or is happening, the more focus we need to get on quick models that can have a quick turnaround. 
Jon Krohn: 00:14:07
Nice. In some games that could include things like the bad guys. 
Carly Taylor: 00:14:11
Yes. 
Jon Krohn: 00:14:13
They need to be fully automated but potentially operating online. Yeah, that’s super cool. I’m so out of the loop on fun new video games. I used to be really big into Grand Theft Auto. 
Carly Taylor: 00:14:28
Oh, my gosh. Grand Theft Auto’s so fun.
Jon Krohn: 00:14:32
What is [inaudible 00:14:33]? 
Carly Taylor: 00:14:32
If you liked Grand Theft Auto for the just like runaround, because that’s what I always did. I don’t think I ever did missions. 
Jon Krohn: 00:14:37
Yeah, yeah. 
Carly Taylor: 00:14:41
I just liked to go around and mess things up. 
Jon Krohn: 00:14:43
Oh, I like just leaving everything pristine. So, yeah, yeah, yeah.
Carly Taylor: 00:14:48
That’s a little crazy. 
Jon Krohn: 00:14:49
I just drive around, follow the road rules. It would drive me crazy that I couldn’t signal. 
Carly Taylor: 00:14:57
Oh, my god. 
Jon Krohn: 00:14:57
Like I couldn’t put on the right hand or the left hand signal in Grand Theft Auto, but the bots can, and they expect that behavior from you. So, when you don’t put your turn… If you just turn without putting your turning signal, people will be… Like pedestrians will be surprised, people will honk at you, and I’m like, “I can’t do it. They’re broken.” 
Carly Taylor: 00:15:14
Yeah, and see here I am. I’m just running people over, rear-ending people, you know? You try to get to five stars, that’s the whole point of the game. Or, at least that’s what I thought. 
Jon Krohn: 00:15:23
No, no. 
Carly Taylor: 00:15:23
Most wanted. 
Jon Krohn: 00:15:26
I’m just trying to… like I’ll go buy a new suit. 
Carly Taylor: 00:15:29
You need to play trucking simulator where you can drive a truck through America. Follow the road rules. 
Jon Krohn: 00:15:37
I’ve played that. That was an early 10 years ago. I worked as a trader at a hedge fund and we got an early VR system and that was one of the two games that we had. It was- 
Carly Taylor: 00:15:48
Really? 
Jon Krohn: 00:15:49
… real life trucking. 
Carly Taylor: 00:15:51
Yeah. 
Jon Krohn: 00:15:51
That’s the kind of crazy stuff that you get into on a trading floor in New York. 
Carly Taylor: 00:15:55
Yeah, I was going to say that. It is not what I would’ve thought the traders were getting into, but nothing surprises me anymore. 
Jon Krohn: 00:16:02
I couldn’t, for the life of me, you’d have to back a truck into… It was pretty easy driving on the highway, but then you’d always, you have to make a delivery and you have to back in to a warehouse. I can’t do that. 
Carly Taylor: 00:16:16
You’re like, “This road can’t handle the weight of my big rig.” Very real world situations. 
Jon Krohn: 00:16:24
Yeah. I don’t know. I like my video games really boring. I like my film and my TV really boring too. The less that happens, the more I enjoy it.
Carly Taylor: 00:16:35
Interesting. Yeah, I feel like you just set me up to shade a whole bunch of TV shows, but I’m not going to do it. 
Jon Krohn: 00:16:40
Yeah, I don’t know. We’re very much off-piste here, but I think it’s because so much of my workday is so intense and there’s all this stuff that has to be done, then I’m just like to be able to watch something and just kind of relax and it’d just be like a conversation. 
Carly Taylor: 00:16:57
Where no decisions are being made and it’s calming. 
Jon Krohn: 00:17:00
Yeah, beautiful cinematography, and that’s kind of what it was with GTA as well. It was like I can just drive around and enjoy different kinds of environments. 
Carly Taylor: 00:17:09
Okay, well I was going to recommend you play Just Cause next, which is the opposite. So, now I’m going to revoke that recommendation because if you want straight chaos, that’s the game for you. 
Jon Krohn: 00:17:21
Just Cause. I haven’t even heard of that. 
Carly Taylor: 00:17:24
It’s literally… It’s just because why would you blow this helicopter up or fly it into the water supply. 
Jon Krohn: 00:17:34
That’s a real game? You can get it on like… 
Carly Taylor: 00:17:36
Just Cause. Yeah, it’s great. Super fun. 
Jon Krohn: 00:17:39
There’s a great- 
Carly Taylor: 00:17:40
It’s for the chaotic people like me in the world who- 
Jon Krohn: 00:17:43
Nice. 
Carly Taylor: 00:17:44
… to relax, I listen to heavy metal, and then play Just Cause. Blow helicopters. 
Jon Krohn: 00:17:49
Interestingly, I do listen to metal, so I don’t know. 
Carly Taylor: 00:17:54
You’re just an enigma. 
Jon Krohn: 00:17:56
Yeah, I don’t know. Yes, yeah, nobody’s one-dimensional. At least not on the Data Science podcast. I’m sure our listeners are multi-dimensional as well. All right, cool. So, we’ve got some game recommendations, we’ve learned about some new gaming-related careers. What’s the future of gaming, Carly? We’ve got faster and faster processors. More and more data can be stored. More and more people are connected on the internet and playing video games with each other. What’s going to happen next? What’s exciting? 
Carly Taylor: 00:18:28
Yeah, I’ve got a lot of guesses about this that are just founded in what I like to play personally, and what I like to do. 
Jon Krohn: 00:18:36
More destruction. 
Carly Taylor: 00:18:37
Yeah, if I had to take a guess, it’s be things that get blown up better. 
Jon Krohn: 00:18:43
So, something that used to drive me crazy in the few moments where I would get bored and finally start blowing stuff up in Grand Theft Auto. There’s so many kinds of structures that don’t get blown up, you know what I mean? 
Carly Taylor: 00:18:56
Yes. 
Jon Krohn: 00:18:57
That’s disappointing. So, you’re saying that maybe in the future everything will have real physics. 
Carly Taylor: 00:19:02
You know what’s interesting is that idea of persistence is really plaguing to how you would build a game. Let’s say we’re in a place that has snow and we’re walking, how long do you let our footsteps persist in the snow? Decisions like that that will add to the overhead of the game. How persistent is the world for everyone involved. Do you see my footsteps as well, or is it just on my PC that I see them? Those kind of design decisions add latency and overhead and all the kind of things that you try to minimize, and so, yeah, you’re constantly looking for things like that that are like, “Make it cool and realistic, but also don’t blow up the footprint of the game.” 
Jon Krohn: 00:19:52
Wow. Physics being real. Yeah, that’d be super cool. Persistence, that’s super important. Okay, but those don’t sound- 
Carly Taylor: 00:20:03
Like the future of gaming. 
Jon Krohn: 00:20:04
It doesn’t sound super exciting. 
Carly Taylor: 00:20:06
The future of gaming is everyone’s footsteps are always there. You heard it, you’re first folks. 
Jon Krohn: 00:20:11
Wow. 
Carly Taylor: 00:20:13
No, I’d say that. So, for me as a player, that is part of the fun. I was like, “How many people can we get in the same experience? How many people can we really have doing the same thing at once?” So, to me, I see the future of gaming as games getting even bigger, the experience getting even more and more across different people, across different countries, bringing more people together. I also see the future as being a place where different people play games than maybe they do now. Part of my content strategy on LinkedIn has always been to speak to people who maybe don’t see themselves in a game dev role, or maybe don’t see themselves reflected in the standard characters in a video game, because it’s traditionally been this male dominated field, and I think the future of gaming is democratizing that more. 
 
00:21:05
We’re seeing different people playing video games than maybe we thought. We’re seeing mobile gaming becoming like this exploding into the scene. It’s always been big, but I think the future of mobile games is getting even bigger as our phones get more and more capable, and so I think there’s a lot of interesting, exciting areas where this is just going to keep growing, and now video gaming, it’s a bigger industry than Hollywood and the music industry combined, and so the potential is unlimited. 
Jon Krohn: 00:21:37
Yeah, I don’t know if the stat’s offhand, and you might, but when games like Call of Duty come out, they will, in a week, eclipse the top selling blockbuster film of all time.
Carly Taylor: 00:21:50
Yeah, and that’s happening more and more often now, and I think that’s just going to become the norm. I bet now Hollywood is going to benchmark themselves against us. 
Jon Krohn: 00:22:02
Maybe that’s something that they’ll very much try to avoid. 
Carly Taylor: 00:22:05
Yeah. They’re going to be like, “We actually outsold this top video game this week in the box office,” and they’re going to try to flex that way, as opposed the other way around. They’ve always been the measuring stick for entertainment. Not anymore, we’re coming for them. 
Jon Krohn: 00:22:19
Cool. Well, yeah, and interestingly it blends together a lot of like I’m 10 years out of date on stuff like Grand Theft Auto, for example. But, it had such beautiful cinematic sequences in it. 
Carly Taylor: 00:22:33
Just amazing. Yeah. 
Jon Krohn: 00:22:34
I would get moved playing the game, and the music I used to love. So, that was one of the things, I’d be driving around 
Carly Taylor: 00:22:43
Oh, my gosh, there were radio stations in Grand Theft Auto. I need a Spotify playlist that’s just Grand Theft Auto radio. 
Jon Krohn: 00:22:50
They have them. 
Carly Taylor: 00:22:51
What? 
Jon Krohn: 00:22:52
Yeah, Spotify has them. Spotify has the radio stations from… So, for me, I think it was GTA 4 was the big one for me, 10 years ago. 
Carly Taylor: 00:23:00
For me, it was Vice City. 
Jon Krohn: 00:23:02
Yeah, there you go. I bet they have that one too, but- 
Carly Taylor: 00:23:04
That’s great. 
Jon Krohn: 00:23:05
Yeah, Spotify has those for sure. 
Carly Taylor: 00:23:07
That’s awesome. After you listen to this episode, everyone go find the Grand Theft Auto playlist on Spotify. 
Jon Krohn: 00:23:13
Yeah. I’ll put it in the show notes. 
Carly Taylor: 00:23:15
That’s great. Here’s some music to listen to while you code. 
Jon Krohn: 00:23:20
For sure. It would also it would open my eyes to other genres because somehow in a virtual experience, I am more open to being a different kind of person than I am- 
Carly Taylor: 00:23:33
I love that. 
Jon Krohn: 00:23:34
… in real life. So, I don’t typically listen to hip hop, but I would listen to it in Grand Theft Auto and be like, “Oh, wow. Some of this is amazing. I love this.” It would broaden my vistas. There’s this specific jazz song that I’ll dig up and put in the show notes. 
Carly Taylor: 00:23:53
You should. That’s great. 
Jon Krohn: 00:23:54
That is one of my favorite songs ever. I think it’s called Songs of Innocence or maybe that’s the album that it’s from. But, I absolutely loved the song. 
Carly Taylor: 00:24:02
I can’t wait to listen to this. 
Jon Krohn: 00:24:06
Yeah. I don’t know, there’s probably something interesting there around the psychology of when you’re in that virtual environment, it’s like it gives you carte blanche to change your habits and experiment with different ways of being and thinking than maybe you wouldn’t in the real world. 
Carly Taylor: 00:24:31
I totally agree. Yeah. I think it opens up different horizons and perspectives to people. You meet people from everywhere, I think it’s just anyone that says video games are always negative, I think has probably not played enough, because I’ve met some of my best friends playing games, and I know a lot of people who have too. I was just at a wedding where the best man and the groom met playing Call of Duty. 
Jon Krohn: 00:24:56
Wow. So, they’ll like meet by chance? They’ll be assigned on the same team or something- 
Carly Taylor: 00:25:03
Yeah. And you’re just chatting. 
Jon Krohn: 00:25:05
… and then you wear your headset and you chat, and then you can in the future meet up with them again. You’re like, “Okay, let’s meet again.” 
Carly Taylor: 00:25:11
Yeah, you can add them as a friend, and hang out. Yeah. 
Jon Krohn: 00:25:13
Cool. Are there spaces in Call of Duty for just like hanging out? 
Carly Taylor: 00:25:18
I, honestly, what I do with my friends is we get on the headset and then we all edit our weapons and our load outs, and so we just chat while we all play around with our guns and sometimes we don’t even get into a match. 
Jon Krohn: 00:25:31
Oh, really? So, that’s like- 
Carly Taylor: 00:25:32
We just talk about like, “Oh, I just got this scope, is this good?” And like give each other recommendations and shoot the sh**. Yeah, it’s fun. 
Jon Krohn: 00:25:39
Cool. 
Carly Taylor: 00:25:40
Yeah. 
Jon Krohn: 00:25:42
So, then what do you think about the emergence of… In some ways to you, it must seem comical that there’s this supposed separate thing called a metaverse when what you are just like this is like, “How is Call of Duty not already that?” 
Carly Taylor: 00:25:58
I think that at least the people that I follow that talk a lot about the Metaverse would argue that online gaming is already the metaverse. I think that the building blocks, like I said, of online engineering and low-latency machine learning are going to be the future careers and the future of what else we think the metaverse might be. But, when I think of the Metaverse, that’s exactly what I think of is just a bunch of people online experiencing something together. It doesn’t have to be this dystopic vision that I think people think of when they think of maybe what Facebook is maybe trying to push where it’s overrun with advertisements and it’s a way for companies to kind of monetize or monopolize. 
 
00:26:45
I think that the future of the Metaverse, at least what I hope it ends up being as a consumer is a place the internet is now, where it’s an open forum and hopefully it continues to be a little bit more democratized where there’s not one company that owns the internet, even though Google’s kind of getting there, but they still don’t, right? You can avoid Google and still go on the internet if you want, and I hope that the Metaverse is the same way where it’s a joint collaborative effort with a bunch of different companies and a bunch of different places for people to experiment. Maybe people make their own content the same way we do on LinkedIn. 
Jon Krohn: 00:27:25
Got it. 
Carly Taylor: 00:27:25
So, a little bit less of just a place to get served ads. 
Jon Krohn: 00:27:29
Got it, got it, got it. So, the metaverse would be like it would have some kinds of open standards that would allow interoperability. So, somebody could be in Call of Duty and then walk through a door and all of a sudden they’re in some other companies. Like it’s not an Activision product, but from this Activision product, you can interact outside with completely different companies or charities, open source groups. 
Carly Taylor: 00:28:03
Yeah, whatever you want, right? It’d be kind of a better version of what the internet is now, I would hope, where we can get more connected and not less connected. 
Jon Krohn: 00:28:15
Cool. Are there specific games out there that you would recommend for listeners if they’re just into this idea of socializing online in a virtual environment? 
Carly Taylor: 00:28:27
I think that there are some probably second life type of games that you could play where you can socialize. I honestly don’t play them very much. Most of my socializing is done in Call of Duty, I’m going to be totally honest. I also like to play solo games as well, so I may be on the opposite spectrum of someone who likes to decompress and be alone while I play games too, and so I go back and forth. 
Jon Krohn: 00:28:52
Oh, me for sure. That was always like, “Yeah, it’s kind of nice to have.” Especially after a workday of where every decision matters. 
Carly Taylor: 00:29:01
Yes. 
Jon Krohn: 00:29:01
It’s so nice to be able to go into a place where it just doesn’t matter. 
Carly Taylor: 00:29:04
Yes, yes. 
Jon Krohn: 00:29:05
Like you’re not impacting anybody, it’s just bits flipping in the box underneath my TV. 
Carly Taylor: 00:29:11
Playing like Animal Crossing and it’s just very positive, happy experience. There’s no critical decisions, we’re talking about turnips. 
Jon Krohn: 00:29:25
That does sound great. 
Carly Taylor: 00:29:27
You’re like, “Sign me up for turnips, I’m here for it.” 
Jon Krohn: 00:29:33
Mathematics forms the core of data science and machine learning, and now with my mathematical foundations of machine learning course, you can get a firm grasp of that math, particularly the essential linear algebra and calculus. You can get all the lectures for free on my YouTube channel, but if you don’t mind paying a typically small amount for the Udemy version, you get everything from YouTube plus fully worked solutions to exercises and an official course completion certificate. As countless guests on the show have emphasized, to be the best data scientist you can be, you’ve got to know the underlying math. So, check out the links to my mathematical foundations and machine learning course in the show notes or at jonkrohn.com/udemy. That’s Jon Krohn.com/udemy. 
 
00:30:17
So, all right, so that was a very interesting foray into video games without really much data science talk at all. So, let’s bring it back to that a little bit. What kinds of tools or software packages do you use regularly or are excited about that our listeners should hear about? 
Carly Taylor: 00:30:35
Yeah, so I actually just learned that this is one of Serg’s favorite packages, but I love the SHAP package in Python for using Shapley values to do explainable modeling. I think that as a data scientist, one of the best things you can do for yourself, for your stakeholders, for the people you’re mentoring is to make everything you do more explainable, and I think that starts with having a well-defined problem and a well-defined problem space, a well-defined solution of what done means for whatever project you’re working on, and then being able to walk even non-technical audiences through what exactly you’re doing, how your solution is going to help solve their problem, and what the results of what you’ve done mean. 
 
00:31:26
If they’re asking you why a player is churning out of a game, let’s say, you need to be able to explain, “Okay, this is what the model saw and this is what I think in real terms this means for the business, and what the model’s telling us all these features when they interact, what that means.” I think that can help every data scientist become more effective and be seen as a more trusted collaborative partner, as opposed to this person who knows this kind of magic like Python and can build machine learning and no one knows what that means, and it’s all a big mystery. When you’re early in your career, you might fall into the trap of feeling like that gives you power, that you know this thing that other people don’t, and it does in a sense, it very much does. 
00:32:14
But, what gives you the ultimate power is making other people feel like they can understand what you do and empowering them to think, “Hey, I can actually understand data science. It doesn’t have to be this nebulous thing, it can be something that I can use in my workflow,” and you can kind of open it up a little bit, and I think that will take people to the next level, at least in their careers. That’s what’s worked for me. 
Jon Krohn: 00:32:38
It’s become almost a cliche on this show that when I ask people, “What do you look for in people that you hire?” That is the number one thing that people say is communication. 
Carly Taylor: 00:32:46
Really. 
Jon Krohn: 00:32:47
Yeah. 
Carly Taylor: 00:32:47
Yeah. 
Jon Krohn: 00:32:47
The ability to… Great that you have all this technical background, but being able to communicate that to other people on the data science team, or outside of the data science team, to other people in the company, to external stakeholders, being able to convey in a way that people can understand and show the impact of what you’re doing is the most desired skill that we look for, and actually, I’m going to ask you that question right now, but- 
Carly Taylor: 00:33:12
Well, you just answered it. 
Jon Krohn: 00:33:13
There we go. Wait, wait, wait. Just before we get to that, and I ask you what you look for in people that you hire, I just want to also make the point that you, before the program when we were talking about SHAP, you described specific functionality that I thought was really cool that I hadn’t thought about using that package for, and it’s kind of obvious to me now, which is just for plotting interaction terms in a way that you can easily see how all the different interaction terms in your model are interacting with each other, and having their impact on your outcome variable. 
Carly Taylor: 00:33:53
Yes. I love that because a lot of us fall into the trap of just looking at feature importance values and you get some sort of magnitude that means nothing other than it’s relative magnitude to another term, but it doesn’t really tell you anything about, let’s say, for credit card fraud, one of the most important features could be the distance that someone from their home is making a charge, and it might seem obvious to us from intuition that for the model, it’s probably the larger, the distance that you are from your normal home, the more likely a charge is to be fraud. But, you can’t say that definitively just by looking at a feature [inaudible 00:34:34], right? It could be that the closer you are to home, the more likely that a credit card transaction is fraudulent.
 
00:34:41
So, without some sort of directional explanation, you kind of don’t know that, and that there are situations where it becomes more complicated, where it might be useful to understand not just the magnitude, but the direction that a feature has on the outcome and on the prediction that your model’s making. 
Jon Krohn: 00:35:00
Yeah, I think also one of the key points you were making was that the interaction between multiple- 
Carly Taylor: 00:35:06
Oh, for sure. 
Jon Krohn: 00:35:06
… or independent variables with each other, right? 
Carly Taylor: 00:35:08
Yeah. 
Jon Krohn: 00:35:08
So, distance is only a factor when also some other thing is going on. 
Carly Taylor: 00:35:14
Exactly. 
Jon Krohn: 00:35:14
Like they’re not using a chip in pin card or something. 
Carly Taylor: 00:35:17
Yeah. Yes, exactly. 
Jon Krohn: 00:35:20
Cool. All right, so then we can go to the question that I would just [inaudible 00:35:26]. Other than communication, which everyone knows is super important, what else do you look for in people that you hire? 
Carly Taylor: 00:35:34
I would say I look for people who love solving problems, because at the end of the day, what we do is we try to solve problems for other people, we solve problems for ourselves. We’re constantly faced with new challenges and new problems, and if you’re the kind of person who likes to put puzzles together in your spare time, or loves when your family asks you for IT help, because when stuff breaks, you’re ready and you want to figure out what’s going on and you’re curious. I think that those are the kinds of people that I have found are the most capable and excited to get to the root cause of something and find out what’s going on. 
Jon Krohn: 00:36:20
Nice. Yeah, it’s a great answer. Okay, so we’re kind of flip flopping here, and back two years ago when I took over hosting the SuperDataScience Podcast from Kirill Eremenko, one of his main things to me was like, “You don’t want to change gears too many times,” and he’s absolutely right. 
Carly Taylor: 00:36:43
But, we’ve done that a lot, so let’s just keep doing it. 
Jon Krohn: 00:36:46
So, we were just talking about tools and then I had you do talking about career guidance, and I feel like this is kind of just it’s how the conversation’s going, but there’s a really important tool question that I want to get back to. So, changing gears back. Sorry for the jarring topical experience, listener. This is a real life conversation. So, I noticed a post around the time that we started talking about you being on the show, and because you’re so in demand with big Call of Duty releases and these kinds of things, it took a while to get you on the show. So, several months ago, you had a post on LinkedIn where you were answering the question, “Windows or Mac.” So, Carly, what is it, which of the two options do you choose? 
Carly Taylor: 00:37:38
C, none of the above. I know. What a curve ball. No, I will say that for my daily driver at work, I do use a Mac. I find that for data science in particular, I think that Mac has done a lot to make the native Z shell is amazing. They make doing data science really straightforward. But, for my personal computer, I will say that I still use Linux. I don’t know if it’s the contrarian in me or what my issue is because I always have technical problems with LinkedIn, I’ve had technical problems in this podcast, but I refuse to give up. I think it’s some sort of refusal to admit defeat, and persistence. Maybe it’s the problem solver in me, right? Things break, my drivers don’t always work, it’s always something, but it gives me something to do, and it’s like my baby now. This operating system I’ve tailor made to myself, and so- 
Jon Krohn: 00:38:42
Oh, yeah.
Carly Taylor: 00:38:43
If you want to spend most of your time debugging issues with drivers and compatibility, get Linux, get with it, let’s all join a chat and complain together, it’ll be great. 
Jon Krohn: 00:38:55
Yeah, I mean, the amount of flexibility is huge. 
Carly Taylor: 00:38:57
It is. 
Jon Krohn: 00:38:58
So, often you’ll see really hardcore programmers who don’t own a mouse and are doing everything in Emax. They’ll be using Linux for sure, and everything is customized, optimized where it’s keyboard only, your key- 
Carly Taylor: 00:39:18
Yes. 
Jon Krohn: 00:39:19
You never need to leave your keyboard and that is something on a Mac, you can’t really do that, you can’t beat that level of hardcore with your programming. 
Carly Taylor: 00:39:26
Yeah, I don’t like the lack of customizability on either Windows or Mac. I do think that it’s getting a little bit better from what I’ve seen just from my usage of Mac at work. But, I will also say if anyone’s listening, who’s wanting a gaming career to get into game development, Windows is still king there. Most game development happens on Windows, and so all three get used. I actually have a PC right here as well, but… 
Jon Krohn: 00:39:54
So, do you game mostly on the PC or do you have a console? 
Carly Taylor: 00:39:58
I also have a console, so I just like I can’t stop. I have everything. So, I have a PS5 as my daily console, which I love. But, I actually have an Xbox Series X as well, and I prefer the controller on Xbox. I think it’s more ergonomic and it’s easier to use. After a long time, the PlayStation controller actually hurts my hands, but I also probably have carpal tunnel from typing all day. 
Jon Krohn: 00:40:24
I noticed that you probably have a really snazzy keyboard. I could hear before we started recording, very strong hammers. It felt like those keys were going- 
Carly Taylor: 00:40:33
Yes. 
Jon Krohn: 00:40:34
… a very far distance. Do you have a special keyboard- 
Carly Taylor: 00:40:36
Yes. 
Jon Krohn: 00:40:36
… you want to tell the listeners about? 
Carly Taylor: 00:40:37
Well, I do. I have this Realforce keyboard that I love. It’s really cool. Matte black, black on black. It’s actually not as clacky as my other keyboard. I think this one has… No, this one I think has the brown switches. So, this actually has blue switches, which is the clackiest keyboard that I own. I brought this to an office one time and all my cube mates told me that I had to take it home. They were like, “It’s too loud.” 
Jon Krohn: 00:41:06
Yes, it is. For our audio only consumers, we have… Carly was just showing me, the entire time that we’ve been recording this episode, there’s been a keyboard on the shelf behind her. That’s the clackiest. So, just tell us again, what’s the one that you’re using regularly? 
Carly Taylor: 00:41:24
So, my regular daily driver, I think, are the MX brown switches, and then this is a blue, which is extremely clacky. So, my display keyboard, that isn’t my daily driver anymore because I can’t be on a video call and also type with that one because everyone knows I’m not paying attention, but with this one I can kind of stealth multitask, and you can hear me typing, but you’re not sure that’s what it is. With this one you’re like, “Are you in a tornado? Is there something going on?” 
Jon Krohn: 00:41:56
Nice. Then, so I’ve never owned a keyboard that’s like that hardcore. What is the draw of such a loud keyboard? It’s a real joy to type on? 
Carly Taylor: 00:42:07
It’s the tactile feedback. I don’t know how to explain it, but it’s the same reason why my Linux PC is a ThinkPad because I think for a laptop they have the best keyboard. There’s just something about when you’re coding or typing, knowing that your keystrokes, like that tactile feedback is very satisfying to me. It’s also a reason why I absolutely hate my Mac keyboard. I think it’s one of the worst. I have one of the old butterfly keyboards, anyone who knows about them. 
Jon Krohn: 00:42:36
Oh, yeah. 
Carly Taylor: 00:42:37
Will comment on this video and be like, “The worst.” 
Jon Krohn: 00:42:39
Yeah, yeah, yeah. 
Carly Taylor: 00:42:40
Because it feels like you’re mushing your fingers into the sand. It’s just… 
Jon Krohn: 00:42:46
Yeah, and they are famously breakable as well. That butterfly one. 
Carly Taylor: 00:42:50
They are. They break all the time. 
Jon Krohn: 00:42:51
Constantly in backend. Yeah. 
Carly Taylor: 00:42:52
Yes. Not a fan. Mac, get it together. Put some blue switches on your MacBook Pro. 
Jon Krohn: 00:43:00
Since, at least, yeah, MacBook Pro, since the M1, which is the one that I’m currently recording with, it does have more of a keyboard feel. 
Carly Taylor: 00:43:10
Do you have an escape key? 
Jon Krohn: 00:43:13
I have an escape key and I don’t have that touch bar thing is gone. 
Carly Taylor: 00:43:17
Oh, that’s so nice. See, sometimes form over function, these kind of decisions, I’m just like, “You need to not do these things.” 
Jon Krohn: 00:43:25
Yeah. It’s a lot less pretty than previous generations, but way more functional. 
Carly Taylor: 00:43:30
Yeah. Yep. I’ll take it any day. I need a full keyboard. Function keys. 
Jon Krohn: 00:43:35
The only thing that really annoys me about it is just I assumed that we were going unilaterally in the direction of everyone being on USB-C. So, all of my gear is all USB-C all the time, and on the latest MacBook Pros, they dropped one of the four USB-C ports to put in an HDMI port, which they hadn’t had for years, and I don’t have any displays anymore that use HDMI. So, now I’ve just lost a port. 
Carly Taylor: 00:44:03
All of my displays are USB-C now. 
Jon Krohn: 00:44:03
I know. 
Carly Taylor: 00:44:07
Of all the ports to choose to bring back, why HDMI? I have an issue with Mac’s lack of ports, but it’s never been about, “I wish I had HDMI.” It’s like, “I wish I had a USB-A port.” Let’s just be real. That’s old school sh** here. Like this keyboard. 
Jon Krohn: 00:44:23
Yeah. The thing that would happen to me, and I have seen happen to other people, is you’d get to an important business meeting at an investor’s office, and all they have is HDMI, and you’re like, “Ah, no, I forgot my adapter.” 
Carly Taylor: 00:44:36
Everyone gather around my computer. 
Jon Krohn: 00:44:39
Yeah, exactly.
Carly Taylor: 00:44:41
That’s [inaudible 00:44:42]. 
Jon Krohn: 00:44:42
Okay, cool. So, some great hardware tips there. 
Carly Taylor: 00:44:45
We’re really doing a good job here. 
Jon Krohn: 00:44:49
All changing gears. So, you have a master’s in computational chemistry. Seriously, that’s where I’m going with the next question. 
Carly Taylor: 00:44:59
Oh, my gosh. Okay, I’m [inaudible 00:45:01]. 
Jon Krohn: 00:45:02
But, I will tie it into the conversation from before by saying how did you transition into data science from that past? I mean, to me, actually having known a fair bit about bioinformatics and doing a lot of computational biology stuff in my graduate studies, I have a really good sense of why computational chemistry was so useful to you in data science. But, yeah, tell us about that transition from computational chemistry to data science. What parts were easy, what parts were hard, where do you have advantages because of that background? 
Carly Taylor: 00:45:37
I would say the easiest part was that I had already… I had the statistics training, I had the mathematics training for the computational part. I had already been using Python, I was also using Fortran, which ended up not being helpful. [inaudible 00:45:51] is anyone out there messing with that. But, there were a lot of technical things that I think lined up really well and made a lot of sense. But, in terms of the hardest thing for me, was coming up with a way to explain to recruiters, hiring managers, why they should hire this chemist, because when they saw my resume, that’s all they saw, was like, “Oh, you’re a chemist, all of your experience is in chemistry, why are you applying to this job? Did you accidentally apply?” Then the questions about Breaking Bad would come up. They’d be like, “The only thing I know about chemistry is about how to make meth.” 
Jon Krohn: 00:46:31
Oh, my god. 
Carly Taylor: 00:46:31
I’m like, “Okay, well, I guess this interview’s already over if that’s what you’re thinking about.” So, for me, it was like crafting that elevator pitch, coming up with a story for why I would be a good addition to their team, what skills I could bring outside of chemistry, and how they translated. That was the hardest part, was like crafting that narrative, crafting that story, making sure my resume told the same story that I would tell in interviews, and everything was cohesive enough that it made sense, and they were like, “Of course a chemist would apply to this job,” and not, “What is this?” 
Jon Krohn: 00:47:07
Right, right, right. So, would that be your general guidance to people moving from a hard quantitative science? I mean, it seems crazy to me, if it’s a computational chemistry, that seems like really obviously related to what we do in data science, but do you have advice for people coming from quantitative sciences, hard sciences? I guess, that would be it, to make sure that you craft your resume in a way so that the story arc is clear to the interviewer just from glancing at your resume that even though your degree says chemistry in the title, that throughout that degree, you were learning how to apply quantitative methods computationally. 
Carly Taylor: 00:47:49
Yes. 
Jon Krohn: 00:47:50
Which, guess what? Is data science. 
Carly Taylor: 00:47:53
Exactly. Yep. I would say that’s the best thing you can do. Get a second, a third, a fourth opinion on your resume. If someone looks at it and is like, “Yeah, I still kind of don’t know what some of these keywords mean. As a chemist, why are you still referencing ultraviolet radiation?” Take the things off your resume that you think tell the story of who you were in your old field, and tailor it to the new one, and this can be really hard for people, and I get it because I was- 
Jon Krohn: 00:48:23
You publish all these papers. You’re like real proud of them. 
Carly Taylor: 00:48:24
Yeah, I wasn’t an expert. Exactly. But, I was more of an expert than most people in my area, and moving to an entirely new career meant I had to let that go and I had to start over, and I couldn’t hold onto that and try to convince people of my expertise in the way I was used to, and I had to switch and completely change the language I used and start over as a newbie. It was really hard, and I think of this, it’s actually it’s not really related at all, but it makes me think of one of my really good friends is her English isn’t her first language, and she said to me once, it was so funny, that sometimes she feels like she just wants to say to people, “You have no idea how smart I am in Spanish.” 
 
00:49:12
When you’re struggling to explain yourself in a language that’s not your native tongue, that you feel like you have all these other skills that maybe people aren’t seeing and they don’t know about, and that is a little bit, not to the extent of that, but that’s a little bit about how I felt when I was leaving behind all these skills I had in chemistry, and I had to just be like, “You have no idea how good I was at this thing, but it doesn’t matter anymore and I have to move on.” 
Jon Krohn: 00:49:36
Nice. That was a great answer. So, with these kinds of career tips that you have, these are the kinds of things that you actually offer to the public at anytime. So, anybody who’s listening to this show, you can go right now to, say, Carly’s LinkedIn page, and at the top of her page there is a button at the very, very, very top where, I don’t know, at some point this year, I think you added LinkedIn out of the capacity to have URLs right at the top of profile. 
Carly Taylor: 00:50:09
Yeah. 
Jon Krohn: 00:50:09
So, it says like, “Carly Taylor, her title, what she does, what she talks about, the hashtags, where she’s based,” and then there’s this thing that says, “Let’s connect,” in blue. It’s clearly a URL, you can click on that, and it takes you to a page where you can book time with Carly to get one-on-one career advice. That’s cool. 
Carly Taylor: 00:50:28
Yep. Thanks. Thanks for the shout out. But, yeah. 
Jon Krohn: 00:50:32
I mean, you’re probably aware of other people that do that, but I’ve never seen anybody else do it. 
Carly Taylor: 00:50:36
Oh, really? Oh, that’s [inaudible 00:50:38]. Well, yeah, I love talking to people, I love getting to know people. I’ve tried to find a way to streamline the way that I’m able to connect as I started to reach out to more and more people and my following on LinkedIn got bigger. I had to figure out a way to optimize this. It was more problem solving. It was like, “I’m getting too many emails that I can’t respond to, because the volume, people want to talk one-on-one with me, but I don’t necessarily know how to go about setting this up and balancing it with the fact that I still have a nine-to-five job.” 
 
00:51:09
So, yeah, I’ve tried really hard to make something that works for people to where they feel like they can get one-on-one advice with me if they want. I’ve tried to keep my prices to the point where it’s just like it makes sense enough for me. It’s not as much money as I make in my day job, but it’s enough to where I can dedicate time to you and your problems and I can sit down and spend time with you, and we can come to a conclusion and I can help you as much as I can. 
Jon Krohn: 00:51:33
Yeah, and it’s the kind of advice, especially if you are the kind of person, and I’m sure there’s lots of listeners out there like this, where you’re looking to get into a data science related career or you’re looking to progress in it and you don’t have friends that do it. 
Carly Taylor: 00:51:49
Yeah. 
Jon Krohn: 00:51:51
We probably don’t have that many parents that came from a data science background that we can ask them for advice, and sometimes, especially if it’s maybe something like a career move to another company, or just something you’re considering, can’t talk to your colleagues about it, so then having Carly as your therapist. 
Carly Taylor: 00:52:14
A lot of it ends up being a little bit of that, I will say. 
Jon Krohn: 00:52:17
I bet. I mean, how is a big career change, it’s intricately linked with the whole rest of your life, and actually the episode that aired just before this one on Friday, episode 644, was with this professor who has spent decades researching the entanglement of career and personal decisions. 
Carly Taylor: 00:52:45
Interesting. Yeah, that’s so cool. 
Jon Krohn: 00:52:48
Yes, she has some advice that people can go back to there and dig into if you haven’t already listened to that episode. But, yeah, and any of your specific career questions, you can talk to Carly about, if you’d like to. So, you describe yourself, I think this is kind of related to your data science career counseling and guidance persona, which by the way, also, I mean that’s how you’ve developed this enormous LinkedIn following, right? Is by providing career advice in general by posting about it. So, at the time of recording, you’re over 70,000 followers, which is crazy. So, in all of this persona, you describe yourself as the rebel data scientist. So, what does that mean? 
Carly Taylor: 00:53:34
That’s so funny. Someone else described me that way and I was like, “This is amazing. I love it.” And I was like, “It wasn’t my intention to be seen that way,” but I think that what I’ve always tried to be, especially on LinkedIn, was authentic to myself, authentic for my followers so that they knew I’m not sugarcoating things, I don’t make fluff pieces, I’m not going to make this career or what I do seem like it’s anything other than the reality of what it is, and I think that that very direct communication style, the no nonsense, no sugarcoating, tell it like it is kind of personality that I have anyway comes across as almost rebellious in this space where we’re often told we have to be a certain way or we have to conform to some sort of idea of what a professional looks like. 
 
00:54:30
I’ve always thought that that was kind of BS. I don’t think that there’s one way you have to be, to be a professional, and so I’m just out here with my tattoos and my nose piercing and just trying to get through life and be exactly who I am, and so people know, you know? 
Jon Krohn: 00:54:50
Nice. 
Carly Taylor: 00:54:51
Show up to meetings in hoodies. But, to be fair, in gaming, that’s pretty chill. 
Jon Krohn: 00:54:59
Well, that’s great. It’s nice to have that background. Yeah. So, yeah, can’t recommend highly enough if you’re looking for an advocate to get promoted, to navigate situations with leadership, to mindfully leverage your strengths to optimize your career, follow Carly and consider booking time with her. So, brilliant. One piece of specific career advice that you might be able to provide all of our listeners on air is related to patents. 
Carly Taylor: 00:55:29
Yes. 
Jon Krohn: 00:55:29
So, you have a patent called real-time analysis of infield collected well fractured data. So, that sounds like… So, this is well fracturing. Yeah, I mean, I guess, I don’t know why I’m going to try to guess what this means. I guess it’s like people have wells restoring water or whatever, and sometimes they break and that’s- 
Carly Taylor: 00:55:51
It’s actually for oil. 
Jon Krohn: 00:55:55
Oh, yeah, of course. Fracking. 
Carly Taylor: 00:55:57
Yes. You know what’s funny is that I thought when I was working in fracking, I was like, “This is the most controversial job I could ever have. I’m a data scientist, I’m working with these oil and gas companies, they’ve got this really bad reputation,” and when I switched to gaming is when I received the most negative feedback from people about just like the haterade and the hater comments went off the chain from gaming and I was like, “You guys know I used to work in fracking, right?” That’s a way more… I could understand, like you have a leg to stand on if you’re going to criticize me for that. But, for the game, I was so surprised by it. 
Jon Krohn: 00:56:32
Why do people… Because of violence in games? 
Carly Taylor: 00:56:36
I think people just have very strong opinions about games, about how they should be designed, about what their experience in gaming should be like, about if women should have a role in gaming, about… There’s a lot of baggage there. 
Jon Krohn: 00:56:50
Oh, my goodness. 
Carly Taylor: 00:56:50
Yeah. 
Jon Krohn: 00:56:51
Geez, wow.
Carly Taylor: 00:56:53
It’s not often, I would say, with the majority of my communications with players, everyone’s very happy, they love the game, they’re very complimentary. But, you do have some people that don’t vibe with… 
Jon Krohn: 00:57:06
Rebel datasets. 
Carly Taylor: 00:57:07
Yes. Making fun of trolls is one of my favorite pastimes though, I will say. 
Jon Krohn: 00:57:15
All right, so this fracking patent. So, you have experience with getting patents. What do people need to do? If somebody has an idea and they’re like, “Should I patent it? How do I do that?” How they make those decisions? 
Carly Taylor: 00:57:33
Yeah, I would start by saying, if you have an idea personally and it’s not at work and you think it’s patentable, start doing some research. So, you need to go out there and see what has already been patented, is there something like this that already exists. I will say that just from… First of all, I’m not an attorney. Second of all, I’m not your attorney, so this is not legal advice. But, from just my experience with this, the scope of patents is supposed to be very tight, and so I couldn’t patent, let’s say, a device to talk to other people. That would be way too broad, it’s like, “Are these headphones, is that a speaker, is it a megaphone. What are you talking about?” So, you need to be very precise in what you’re patenting and it needs to have very defined boundaries within the space, to be considered original IP. 
 
00:58:28
Now, if you do still think that your idea is defined enough that it’s patentable and that there’s no prior art in the field that would exclude you from patenting it because someone would beat you to it, talk to an attorney. So, there are patent attorneys, they make very good money, so be prepared to pay. If you are looking for a career and you’re unsure if you want to be a data scientist, but you’ve always liked law, I would say go get a JD and do data science patent law, because you’re going to have tons of people coming to you, paying you the big bucks to understand what’s going on. 
 
00:59:02
But, you’re going to talk to an attorney, you’re going to explain your invention to them, they’re going to ask lots of questions, you’re going to go back and forth with them a lot, they’re going to give you their expert advice on whether they think your invention is patentable, what route you might take to patent it, what they think the scope of the invention should be, and then, yeah, they will, when you pay them, help you go through the entire process with the patent office, go through revisions if necessary, you’ll go through… Your patent can be either non-provisional or provisional, and they’ll walk you through everything and explain to you exactly how that’s going to go down. 
Jon Krohn: 00:59:40
Nice. Yeah, that’s great general guidance, and you started all that off by saying if this is something outside of your work. So, I guess if it’s inside of your work, then you just talk to somebody there and see if there’s general counsel or something. 
Carly Taylor: 00:59:56
Exactly. Hopefully, you have internal counsel at your company who can help you, who’s done this before. So, most, I would say, big tech companies have existing patents, they know exactly how to go through this, they might even have a bounty program for patents. Some people I know will get paid bonuses for patenting because it makes the company look good, it’s a way to protect their IP, it is going to be work from you, so they want to give an incentive program, but it does also look good on your resume. So, even if you’re not getting anything from your employer for doing it, I would say it’s worth asking because you should be able to tell the world about something cool you did, put it on your resume that you have some sort of machine learning patent, and it’s an exciting thing. Yeah. 
Jon Krohn: 01:00:42
Cool. All right, so changing gears. 
Carly Taylor: 01:00:48
No surprise. 
Jon Krohn: 01:00:49
We have, yeah, great patent advice there, some great general career advice. Let’s see if you can give some advice to some of our listeners who’ve asked you specific questions. So, when I posted that you were going to be on this show, it was an extremely popular post, more than 25,000 impressions, several hundred reactions, several dozen comments, and some of them are just saying things like Carly’s amazing. 
Carly Taylor: 01:01:19
Agree. 
Jon Krohn: 01:01:22
Some people used seemingly video game expressions that I don’t understand, like, “Security strategy sounds pog.” 
Carly Taylor: 01:01:31
Pog. I’ll let you look that up. 
Jon Krohn: 01:01:39
Okay. Did I just swear on air? I don’t even know. 
Carly Taylor: 01:01:43
No. 
Jon Krohn: 01:01:47
Then we had some really thoughtful questions from people. Viraj Rana had ones about your career, but I think we’ve actually basically covered those. Like why did you transition? Actually, you know what? We didn’t cover that. We talked about how you transition from computational chemistry to data science. We didn’t talk about why. 
Carly Taylor: 01:02:05
Oh, well, this is a great story. So, be me, be 22 when I finished undergrad, and looking for jobs as a chemist, you’ve been told your whole career since high school that chemistry is this viable field, there’s tons of jobs, there’s just jobs growing on job trees everywhere, you can work wherever you want. So, you’re looking and you’re like feeling like this is not the truth, there’s not that many jobs, and you realize that the one job that keeps coming up and the job that nobody wants is working at a drug testing facility, testing people’s urine. Then you realize, “Oh, great. So, the jobs that are so abundant are in urinalysis labs.” 
Jon Krohn: 01:02:53
Oh, wow. 
Carly Taylor: 01:02:53
And this is going to be the job that I could get working in a lab making $15 an hour, and then you go to one of these jobs to take an interview because you’re desperate and you need money, and the interview is you standing in a closet with pee samples for 10 minutes, and if you can last, you get the job, and you realize, “I’ve made a terrible mistake.” 
Jon Krohn: 01:03:18
Wow. 
Carly Taylor: 01:03:20
Yes, and so I loved chemistry, but I became disillusioned with the fact that I had spent from undergrad four years of my life after grad school and additional two, becoming an expert in a field and something that I thought was super useful. I knew how to do everything from making aspirin to telling you if someone had been doing cocaine from looking at their urine. Things, skills that you would think people might need in this world, I could make detergents, I could make you a flavor that tasted butterscotch in a lab. All sorts of cool sh** that you think is going to be really useful. But, I couldn’t find a job, and the jobs I could find were not that great, and my job prospects were making $15 an hour. 
 
01:04:10
So, with all of that combined, and then the fact that a bunch of my friends were getting poached by FinTech companies to do quantitative trading and data science, I realized I actually had the skills to also go do that and I didn’t have to be trapped in a lab with pee for the rest of my life. 
Jon Krohn: 01:04:28
Well, that why all makes perfect sense, and I wonder if the person who asked the question, Viraj, is a computational researcher at Penn State working on protein science, and so, yeah, maybe you too, Viraj, will discover. 
Carly Taylor: 01:04:46
Yeah, it sounds like they found a good career path for themselves. Maybe I just was in the wrong city. 
Jon Krohn: 01:04:51
I don’t know. Yeah, there’s more stuff here about people. I guess, speaking to that, your rebel data science moniker or there’s people like Carly always has such valuable tips and info, and in a no nonsense way too. All right, and then we have a question here from Yousuf Ali who’s a data scientist. I’m not sure if it’s at University of Michigan or whether he went to University of Michigan, I’m just reading the top snippet that we have on Yusuf here that I can see on the comment, and so first of all, Yusuf said that you and I are two of his favorite data science content creators. So, thanks for that, Yousuf. 
Carly Taylor: 01:05:29
Thanks, Yousuf. 
Jon Krohn: 01:05:31
So, he couldn’t wait for this episode. I hope you’re enjoying it so far. It has been a lot of laughs, hasn’t it, Yousuf? 
Carly Taylor: 01:05:36
It’s been a riot. 
Jon Krohn: 01:05:39
So, Yousuf asks, how much of a boys’ club is data science compared to other professions? 
Carly Taylor: 01:05:45
That’s a great question. I’ve had a career trajectory that I would say has been in fields that were predominantly male dominated, starting with STEM and chemistry, moving into data. Now, I’m in gaming. I will say though that for data science specifically, I found that there is a really good balance of men and women in the field. I think that we have some amazing data influencers like Daliana Liu, we’ve got Jess Ramos, we have Megan Lieu, like people out here who are women in data speaking about their experiences, doing kick-ass stuff in the field, and really making a name for themselves. 
 
01:06:31
So, I don’t think that data science in particular has the same issue that some of the other industries I’ve worked in. Oil and gas too, we were just talking about that, right? That was a total boys’ club, and so I think that data is actually a good mix of men and women and non-binary individuals who are all just coming together, looking at data problems, trying to make sure that we’re using data and AI ethically, we’re talking about the cool stuff we’re building, we’re bringing other women and underrepresented groups to the table, and so yeah, I’d say that it’s been more balanced than other industries in my experience. 
Jon Krohn: 01:07:09
Nice. I’m glad that you’re having that experience. 
Carly Taylor: 01:07:10
Yeah. Cute.
 
Jon Krohn: 01:07:12
Yeah, that was a nice answer. Then, yeah, so we have lots. There was so many questions and comments. Many of the questions, in fact, I’m going to say of all the other questions that people asked, we’ve either already covered them or they were asking for stuff that I am certain would be something that you can’t talk about on anyway because they’re really way too specific about proprietary stuff that you’re doing at work. So, great questions, love the level of engagement, and so thank you everyone for those questions and hope you have enjoyed this episode. All right, Carly, we are actually wrapping up.
Carly Taylor: 01:07:49
Wow. 
Jon Krohn: 01:07:50
Which means that it’s time to ask you for a book recommendation if you have one. 
Carly Taylor: 01:07:54
Okay. I’m turning around for everyone who’s just listening and looking at my bookshelf right here. You know what I’m going to have to do? 
Jon Krohn: 01:08:03
We’re lucky that you have a mic that is on a headset so you can be turning around and doing whatever, and- 
Carly Taylor: 01:08:10
I know, except- 
Jon Krohn: 01:08:10
… your [inaudible 01:08:11] looks great. 
Carly Taylor: 01:08:11
Except the cord’s not quite long enough for me to get very far from my computer. 
Jon Krohn: 01:08:14
So, you couldn’t pick the book that you wanted, you could pick the book that you could reach. 
Carly Taylor: 01:08:20
I would never do that to my friend, Joe. Okay, so it looks like this is going to be mirrored on video, but the fundamentals of data engineering, I’ve really good friend, Joe, wrote this with his partner, Matt. This is an amazing book. It came out just this year published by O’Reilly, Fundamentals of Data Engineering: How to Plan and Build Robust Data Systems. I would say that anyone who is in data needs to read this book because it’s not just for data engineers. Understanding how the data that you need gets to you, how to make smart decisions about your data, how to make smart decisions about your infrastructure, or even how to talk to those geniuses who build it for you. If you don’t have to do it yourself, this right here, highly, highly, highly recommend, and the bird on the front is so cute. 
Jon Krohn: 01:09:11
I also highly recommend that book. We had Joe Reis and Matt Housley on the show specifically talking about that book in- 
Carly Taylor: 01:09:18
Oh, I love that. 
Jon Krohn: 01:09:19
… episode number 595, and I implore constantly on the show how, as the orders of magnitude of the datasets that we’re working with increase year over year, being able to engineer your own data pipelines, as a data scientist, is becoming an increasingly essential part of the job. You can’t just rely on having the nice, tidy, cleaned up, relatively small dataset that you can work with and build a machine learning model from, as a data scientist anymore. You need to, today, and even more so in the future, be able to engineer your own data pipelines and get those data for your own models. 
Carly Taylor: 01:10:03
Yes, I completely agree. What a way to be a competitive candidate too, right? Like a data scientist who understands the fundamentals of data engineering. Can’t get any better than that except for someone who can also communicate. 
Jon Krohn: 01:10:17
For sure, and who loves solving problems. So, Carly, it’s been an amazing episode. I wish it could go on forever, but sadly, even good things must come to an end, and so my final question for you is how should people be following you after the show? We know that you have this hugely dominant presence on LinkedIn. We’ve already talked about that on air. Is there anywhere else that people should be following you? 
Carly Taylor: 01:10:45
I have some new things coming in 2023. I will just say, expanding to new platforms. So, I will be updating my LinkedIn following with that new news coming next year. 
Jon Krohn: 01:10:57
All right, and by next year, well, that’s early 2023, right? 
Carly Taylor: 01:11:03
In three weeks. 
Jon Krohn: 01:11:06
Yeah. So, we’re recording this episode right at the end of 2022, but it will be published in early 2023. So, early this year. 
Carly Taylor: 01:11:14
Early this year. Maybe it already happened. I don’t know. Are we living in the future? 
Jon Krohn: 01:11:24
“Living in the future.” That was a relatively niche reference to the South Park movie that was recently released on Paramount. 
Carly Taylor: 01:11:31
Will watch it. 
Jon Krohn: 01:11:33
It’s so funny. I really, really like it. 
Carly Taylor: 01:11:34
Is it good? I need to go watch it actually. 
Jon Krohn: 01:11:36
It’s literally the only reason to get a Paramount plus subscription and that point is made many times in the South Park movie. So, that’s what I did. I watched it- 
Carly Taylor: 01:11:43
Is it? I love that [inaudible 01:11:45] movie. 
Jon Krohn: 01:11:46
… and then ended my subscription. All right, so Carly, thanks so much for being on the show, and I’d love to check in again in a couple of years, see how you’re doing, this was such a fun episode. 
Carly Taylor: 01:11:58
Yes, let’s do it. I’m always around. I’ll be in this room. 
Jon Krohn: 01:12:07
Nice, nice. That was so much fun. Hope you had a blast too. In today’s episode, Carly filled us in on how online engineers use SQL, PySpark, and Kafka Streams to enable real-time machine learning within a low-latency gaming experience. She also talked about how the video games of the future will have real physics everywhere, more persistence, more simultaneous players, more demographics represented, and more of a presence in mobile operating systems. She also talked about how she’s excited about SHAP for explainable AI, particularly for plotting interaction terms. She talked about why the super customizable Linux operating system is her favorite operating system in general, but that she uses a Mac, particularly its Z shell for work and data science. She filled us in on how to go about filing a patent, whether you’re part of a big company or you’re out on your own. 
 
01:12:59
As always, you can get all the show notes including the transcript for this episode, the video recording, any materials mentioned on the show, the URLs for Carly’s social media profiles, as well as my own social media profiles at www.superdatascience.com/645. That’s www.superdatascience.com/645. If you too would like to ask questions of future guests of the show, like several audience members did during today’s episode, then consider following me on LinkedIn or Twitter as that’s where I post who upcoming guests are, and ask you to provide your inquiries for them.
 
01:13:28
Another way we can interact is coming up on March 1st when I’ll be hosting a virtual conference on natural language processing with large language models like BERT and the GPT series architectures. It’ll be interactive, practical, and it’ll feature some of the most influential scientists and instructors in the large natural language model space as speakers. It’ll be live in the O’Reilly platform, which many employers and universities provide access to. Otherwise, you can grab a free trial, a free 30-day trial of O’Reilly using our special code, this is brand new, first time I’m saying it on air, we just got a special code from O’Reilly to get you this free 30-day trial, and the special code is SDSPOD23. That’s SDSPOD23, and we’ve got a link to that code ready for you in the show notes to click on.
01:14:17
All right, thanks to my colleagues at Nebula for supporting me while I create content like this SuperDataScience Podcast episode for you, and thanks of course to Ivana, Mario, Natalie, Serg, Sylvia, Zara, and Kirill on the SuperDataScience team for producing another outstanding episode for us today. For enabling that super team to create this free podcast for you, we are deeply grateful to our sponsors whom I’ve hand selected as partners because I expect their products to be genuinely of interest to you. Please consider supporting this free show by checking out our sponsors’ links, which you can find in the show notes. If you yourself are interested in sponsoring an episode, you can get the details on how by making your way to jonkrohn.com/podcast. Last but not least, thanks to you for listening all the way to the end of the show. Until next time, my friend, keep on rocking it out there and I’m looking forward to enjoying another round of the SuperDataScience Podcast with you very soon. 
Show All

Share on

Related Podcasts