Jon Krohn: 00:00:00
This is episode number 521 with Khuyen Tran, technical Writer for NVIDIA and Towards Data Science, and data science Intern at Ocelot Consulting.
Jon Krohn: 00:00:14
Welcome to the SuperDataScience podcast. My name is Jon Krohn, a chief data scientist and best-selling author on deep learning. Each week, we bring you inspiring people and ideas to help you build a successful career in data science. Thanks for being here today, and now let’s make the complex simple.
Jon Krohn: 00:00:43
Welcome back to the SuperDataScience podcast. Today’s guest is the remarkable and inspiring Khuyen Tran. Khuyen is a force to be reckoned with when it comes to communicating data science. Since becoming an author for the Towards Data Science blog less than two years ago, her articles garner a staggering 100,000 views per month. The quality and the reach of Khuyen’s writing recently led the microchip manufacturing giant, NVIDIA, to ask her to be a technical writer on their developer blog as well. In addition to this long-form blogging, she also writes practical daily posts featuring Python code on her LinkedIn and her own Data Science Simplified, a short-form blog, leading to her developing a highly engaged and quickly-growing social media following of over 25,000 subscribers in only a little over a year. The craziest thing about Khuyen’s meteoric rise in such a short time is that she’s still an undergraduate student. She currently has a perfect 4.0-grade point average in the computational and applied mathematics degree that she expects to complete next year.
Jon Krohn: 00:01:55
In today’s episode, Khuyen fills us in on how publishing your writing can skyrocket your career. She knows, from experience, her own writing has led to her landing four data science jobs, including her current role building freight transport models for Ocelot Consulting. She also fills us in on her tricks for maximizing engagement with the content you publish, her favorite data science tools and approaches, and her tricks for prioritizing and being as epically productive as she is across her studies, her data science work and her prodigious technical writing. Today’s episode will appeal most to you if you’re an early-career data scientist, or if you’re someone at any stage of your profession, and you’re keen to learn how you can dramatically accelerate your career by sharing your voice online. All right, are you ready? Let’s do it.
Jon Krohn: 00:02:51
Khuyen, welcome to the podcast. It’s so awesome that you’re here. How are you doing today? Where in the world are you calling in from?
Khuyen Tran: 00:03:00
Hi, Jon. Thank you for having me in the show. I’m in southern Illinois, so it’s very close to St. Louis.
Jon Krohn: 00:03:10
Oh. Close to St. Louis. All right. Because in my head, I guess I always think about Chicago, but of course, Illinois is a big state.
Khuyen Tran: 00:03:19
Right. The funny thing is that everybody, when I say Illinois, they associated me with Chicago, but I’m four hours away from Chicago and half hour away from St. Louis, which is another state.
Jon Krohn: 00:03:34
In Missouri, right?
Khuyen Tran: 00:03:35
Right.
Jon Krohn: 00:03:35
Nice. All right. Do you go down to St. Louis, so that’s the big city, when you want to have a big night out or?
Khuyen Tran: 00:03:42
Mm-hmm (affirmative).
Jon Krohn: 00:03:45
Cool.
Khuyen Tran: 00:03:47
I like St. Louis. It’s not too small, but it is not too big, so you don’t worry too much about traffic jam, and you still have so many things to do in this city.
Jon Krohn: 00:04:00
Nice. Very cool, Khuyen. I know you through your LinkedIn posts. You post to so many LinkedIn posts that are obviously hugely valuable to the LinkedIn community. I can see that basically every single one gets hundreds of reactions, and these are quite practical posts on how to write Python code in particular libraries so that data scientists can make the most of that software. I’ve known about you for a while. I’ve seen these posts for a while, and I’ve had this idea in my mind, “Oh man, I’ve got to see if Khuyen would like to be on the SuperDataScience show.” Then recently at the time of recording, you published a LinkedIn post where you posted a photo of my book, Deep Learning Illustrated, and it got over 500 reactions, which I think is more than any post I’ve ever made. I was like, “You know what? This is it. This is the sign. I just got to ask Khuyen.” I was like, “She obviously knows who I am now, so let’s see if she wants to be on the show.” I was so happy that you said yes.
Khuyen Tran: 00:05:12
Of course. I mean, I have been a fan of your show, so it’s a no-brainer to say yes to your request.
Jon Krohn: 00:05:21
Nice. I was happy to learn that, that you were a fan of the show already. Super cool. Not only do you write all of these LinkedIn posts, which have tens of thousands of views on average, but you actually also write outside of LinkedIn in a longer form. LinkedIn, you’re constrained to these small posts, and I think you do an excellent job with that small format. You often will do a screenshot of a particular chunk of code that illustrates something clearly, and then you explain what you’re doing in that chunk of code. It works really well in the short LinkedIn format, but you also write long-form articles. You’ve been writing for Towards Data Science for a while now. You’ve written over 100 articles for Towards Data Science. That’s led to a lot of eyeballs on your content, so you’re now over 100,000 views per month on Towards Data Science alone for your long-form blog posts. My first question for you, Khuyen, is what kinds of posts or types of posts do you find resonate most with your audience, or what are you most passionate about writing?
Khuyen Tran: 00:06:41
There are two type of posts that I’m very passionate about sharing. You said I write very short posts and long-posts in form of articles. For the articles, because it’s longer, so I take that as an opportunity to use the knowledge that I know, so maybe some tool or some concept I know, and put them together in an article and try to explore what I’m curious about. My article is a mix of what I really know and what I want to learn more about.
Jon Krohn: 00:07:22
Yes. That’s a good… I’m pretty similar. I’m often like, “Wow, there’s something interesting here,” but I really want to dig into this more, and if I commit to writing it… I mean, the book is even a good example.
Khuyen Tran: 00:07:34
I know.
Jon Krohn: 00:07:34
I wrote the book Deep Learning Illustrated because I wanted to really understand these concepts well, so I completely understand what you’re saying.
Khuyen Tran: 00:07:43
I actually heard that from your podcast. I’m like, “Wow, that’s difficult, though.” I’m doing articles. I feel like it’s somehow easier than a book, of course. I was like, “Wow, that’s crazy. I need to…” I saw somebody else doing this, so I probably can do something even more.
Jon Krohn: 00:08:05
No doubt. Actually, I can give you a little bit of a secret, which is that you need to break the work into smaller chunks. So for writing that book, I broke it into a series of videos. I created the content, and learned the content very well for these much smaller deadlines of creating videos. Then once I had… It was about 20 hours of videos done. I was like, “Well, now, all of this will tie together very nicely into a book.” I imagine you could probably do something similar with your blog posts. If there’s some overarching themes around your blog posts, a book could be a combination of them.
Khuyen Tran: 00:08:50
I actually… That’s a very great advice. I had just realized it recently. I have all the LinkedIn posts, and I gathered them into a book called Efficient-
Jon Krohn: 00:09:05
There you go.
Khuyen Tran: 00:09:10
I put together in two days is because… I think it’s still high-quality, because I spent a lot of time per day for each post. Now, I put all those together, so it’s all the research put into one resource.
Jon Krohn: 00:09:32
I just found it here as we were speaking. It’s called Efficient Python Tricks and Tools for Data Scientists. There you go. It looks like there’s a GitHub link to get that. Is it a free book or do you charge for it?
Khuyen Tran: 00:09:45
It’s a free book because-
Jon Krohn: 00:09:46
Oh, wow. There you go.
Khuyen Tran: 00:09:47
I just… Because you see the free resources I already share, so I just want to find a way to put them together into one website where people can play with the code. That makes me happy.
Jon Krohn: 00:10:04
That sounds great. Here I am telling you how you could make a book, and meanwhile, you’ve already done it.
Khuyen Tran: 00:10:09
I actually talked to a lot of people. It is not easy to just come up with decision to write a book. I think it take a lot of motivation to get to write a book.
Jon Krohn: 00:10:24
It was the worst experience of my life, for sure, but now I’m doing it again. I’m writing a second book. There’s something… It’s hugely rewarding when you’re done, but it is complicated because you have to… The whole book… My first book was about 600 pages, and it has to be consistent and correct all the way through. I think when you write a good book, you need to be able to say things. When you’re in chapter four, you say, “Hey, there’s more of this coming up in chapter 14, and this builds on what we’ve done in chapter one.” Somebody who’s reading the book, they’re not necessarily going to read it from front to back. They’ll be referencing back to things, and so there has to be this consistency, this tightness throughout it. I think that’s the hardest part about it.
Khuyen Tran: 00:11:19
I read a lot of technical books. How I read them was I don’t read like… I feel like a book is like a resource that you take some from these books, get some from another book. How I often read book is I just get to the chapter that I’m very interested in, and I did the same with your book. It was very interesting. Actually, I get to one chapter, and from one chapter, I read it consecutively because that’s what I want to learn more about.
Jon Krohn: 00:11:50
Right. That makes sense. Do you remember which chapter it was in mine? Which one was it that you were so interested in?
Khuyen Tran: 00:11:56
That’s a very good question. I don’t have that book with me. I think I put it somewhere, but I think what I was really trying to… I’m using Deep Learning in my current internship. I used a friend book, but I don’t really understand what each parameter do, so reading your book really helps me to understand it because for me… I think it’s showing a lot in my articles. I like to learn from pictures, from graphs. You put a lot of them in your book. It just make it so much easier for me to follow without take a lot of time to understand the math concept behind it. I mean, I’m a-
Jon Krohn: 00:12:50
That [crosstalk 00:12:50].
Khuyen Tran: 00:12:50
I’m a math person. I’m still prefer reading something not too heavy in math.
Jon Krohn: 00:12:57
That was exactly the idea with that is that I think a lot of people can learn concepts more easily visually, and so I was teaching that content in classrooms. I found that there were particular ways that I found of explaining things with a whiteboard using different color markers on a whiteboard. Then I thought, “Okay, this is a good approach for teaching in a book as well.” Then luckily, I was able to partner with Aglae Bassens, who is an amazing artist, and she did an incredible job taking my terrible color sketches, and turning them into beautiful illustrations. Anyway, we’re not here to talk about my books. I’m going to stop, stop, stop. You were talking about your posts. You were talking about how you like to make long-form posts that blend together things that you’re already familiar with and things that you would like to learn even better. You filled us in a bit on the kinds of content that you like to write about, but what are your tricks? How can listeners who might be writing their own posts increase their engagement?
Khuyen Tran: 00:14:08
I have been… So, you are talking about articles or LinkedIn posts?
Jon Krohn: 00:14:14
Well, either, I suppose. I mean, I guess I was more asking about the Towards Data Science articles, these long-form posts, but I guess either could be interesting.
Khuyen Tran: 00:14:23
I could talk about articles first. I have been… There’s not one way that fit all, so I have been playing with different ways of writing. I guess at the beginning, a lot of people… The question I often get from people is, “How can you be consistent with your writing?” My trick was at the beginning, I didn’t write such long article like I do now. I write very short article because it’s very easy to write, and it actually good for the people who don’t have a lot of time to read articles.
Khuyen Tran: 00:15:11
Recently, I wrote longer articles, but there’s one thing that I keep consistent throughout my articles is the interpretability of the articles. What I-
Jon Krohn: 00:15:23
Interpretability.
Khuyen Tran: 00:15:26
Mm-hmm (affirmative). For every of my article, I will give a person who don’t have a technical understanding of the subject, and have them… Maybe they have some statistics knowledge, but not so much about data science. I will have them reading my articles and see if they understand it. And if there’s some point that they don’t understand, I will be like, “Okay, I need to add more pictures here, or I need to explain it better, or I should shorten this article.” Multiple points that I try to optimize in my articles is it’s better to be simple than to be complex. Sometime, people write very long concept. They try to explain something in very lengthy words, but what would be better is to explain the same thing in simpler word, but still deliver the same meaning. Also, I try to put a lot of codes and a lot of pictures between words so that-
Jon Krohn: 00:16:28
A lot of code and a lot of pictures, that helps for sure.
Khuyen Tran: 00:16:28
Yeah, so that people like, “Oh, I can actually apply this for something,” or, “They look very interesting.” It’s like reading a comic book. That’s what I want people to feel like when they read my article. I feel the same vibe when reading a book.
Jon Krohn: 00:16:58
We’re definitely on the same page about that.
Jon Krohn: 00:17:03
This episode is brought to you by SuperDataScience, our online membership platform for learning data science at any level. Yes, the platform is called SuperDataScience. It’s the namesake of this very podcast. In the platform, you’ll discover all of our 50 plus courses, which together provide over 300 hours of content with new courses being added on average once per month. All of that and more you get as part of your membership at SuperDataScience, so don’t hold off. Sign up today at www.www.superdatascience.com. Secure your membership, and take your data science skills to the next level.
Jon Krohn: 00:17:43
As many pictures as I can put in, I agree. It makes things a lot more fun. That was… Years ago, I used to write on Medium as well regularly. That was definitely… I think Medium makes it very easy to insert photos in a way that is engaging. Another nice thing about Medium that you can probably corroborate if people are looking for a place to do their writing, it’s nice because it forces you to have your writing fit into a very specific template that always looks nice. There’s not a lot of options, but you quickly discovered that for most kinds of writing, you don’t need all those options.
Khuyen Tran: 00:18:23
Right. I think Medium only allow you to do H1 case and H2 headline, and that’s it. If you want to do H3, H4, they’re not going to allow you, but that’s the beauty of it, because if they allow you to do more sub-categories, it becomes very messy for the readers.
Jon Krohn: 00:18:50
Exactly. You’ve got heading one, heading two, and that’s it. Exactly. There’s only one way to do bullets. There’s only one way to have numbered lists and so on. Exactly. All right, so you’ve been writing for Towards Data Science for about a year and a half, coming on two years, and already has made such a huge impact, and it’s led to new opportunities. We’re going to talk about a number of those, but one of those new opportunities that has recently come up at the time of recording is you now work for NVIDIA, one of the coolest tech companies in the world, as a tech writer for their developer blog. That’s super cool. I guess that happened as a result of them seeing how popular your Towards Data Science articles are, and maybe even your LinkedIn posts.
Khuyen Tran: 00:19:40
Definitely. They reached out to me because they know me, they know my skills from my articles and LinkedIn posts. It’s an Hispanic effect from my writing, and I’m happy with it.
Jon Krohn: 00:19:56
It’ll expand the impact of your writing. Are there particular… I imagine when you write for Towards Data Science, which actually I should have mentioned right at the outset that Towards Data Science as a Medium publication is one of the most prestigious and well known on the internet not only just on Medium. I should have mentioned that from the start with Towards Data Science. I imagine when you’re writing for Towards Data Science, can you basically write about anything you want? Then when you write for NVIDIA, is that still the same, or are there expectations that you write on particular topics?
Khuyen Tran: 00:20:31
Right. Towards Data Science, even though I say I’m a technical writer for Towards Data Science, but they can have the right to accept my article or reject my article based on its quality.
Jon Krohn: 00:20:46
Right.
Khuyen Tran: 00:20:46
That’s why for every article that I write, I put a lot of effort in delivering high-quality articles. That is the reason why a lot of articles you see from Towards Data Science are high-quality articles.
Jon Krohn: 00:21:03
Right. [crosstalk 00:21:05].
Khuyen Tran: 00:21:07
For Towards Data Science, it’s not like there’s an editor that edit your article because that would be a lot work for them. They have a lot of articles submitted to them every day. But for NVIDIA, I will cooperate with some technical expert to discuss about some tools, or reveal the draft of the article. It’s two different approach. I think both will roll me as a writer.
Jon Krohn: 00:21:38
Totally. You get two different kinds of experiences for Towards Data Science. You have to put extra time into making sure yourself. You have to be the editor, right? At some points, you’re thinking as the writer and you’re being creative, but then at other points, you need to be very critical about what you’re writing and edit it. Whereas with NVIDIA, it sounds like they’ll have some resources that will work with you and be able to edit a bit. That’ll be interesting. That feedback process will no doubt lead to even better writing and just be a great experience, so that’s super cool.
Khuyen Tran: 00:22:19
You asked about the topic, so for Towards Data Science, sometimes I write some topics that not related to data science, but it’s interesting in a way that… For example, I could drive out an article on creating some app using Python to do a particular thing, and they love it, so they’re going to post it. I enjoy working with them because I know that not every article get accepted, and my article’s accepted for most of the time. I’m very happy with… Every time my article is published with them, I’m very happy and proud of myself.
Jon Krohn: 00:22:58
Very cool. We have talked about Towards Data Science obviously and how that recently led to being invited to be a tech writer for the NVIDIA developer blog. I think by the time that this episode is published, you should have probably a couple of those articles out on the NVIDIA blog, so listeners can check that out. An interesting thing here that I want to mention, so most of the stuff that we’ve talked about so far in the episode would be interesting to other people who are interested in writing content themselves, maybe writing LinkedIn posts or writing blog posts or whatever. However, something that we should definitely mention is how doing that can be hugely helpful to your data science career in general. Probably even more broadly speaking, no matter what your career is, if you want to be perceived as a subject matter expert and be recognized by companies in whatever space you work in, writing about that stuff publicly is a hugely useful way to demonstrate your expertise and draw attention, and it can lead to jobs. For example, you, Khuyen, you’ve had four data science internship jobs come out of your writing. You’re currently working at a firm called Ocelot Consulting, and you mentioned to me before the show that you really love working there. I’m keen to learn about what Ocelot Consulting does and what you do there in particular.
Khuyen Tran: 00:24:29
Ocelot Consulting [inaudible 00:24:33] is a consulting firm on different data science practice. It’s about… We advise on cloud computing, on data analysis. Within Ocelot Consulting, we have a smaller startup called Freight Science. As a data science intern, I work on the data science team in Freight Science, and-
Jon Krohn: 00:24:59
In freight science, so that’s shipping trucks or in trains around the U.S.
Khuyen Tran: 00:25:04
Yeah.
Jon Krohn: 00:25:04
Which reminds me St. Louis, Missouri is a big freight hub. Isn’t it? Lots of trains and trucks come through St. Louis.
Khuyen Tran: 00:25:13
It is.
Jon Krohn: 00:25:14
I see.
Khuyen Tran: 00:25:15
You see now, there’s a lot of data science tools for other industry, but there’s not a lot of data science tools out there for trucking industry. What Freight Science do is they help the trucking companies to make the decision that maximize their profit.
Jon Krohn: 00:25:40
Right, so using-
Khuyen Tran: 00:25:41
Right, and we use-
Jon Krohn: 00:25:42
… data and predictive analytics to help minimize freight cost, optimize routes. Sorry.
Khuyen Tran: 00:25:48
Right. It’s actually right. It is two parts, that’s how I think about it is one is to forecast the number of orders, and the second part is based on the number of orders that you think we’ll have on a particular land for a particular company. How are you going to make the decision that optimize the profit of the company in the long-term? That’s two part in it. For me, my role as a data science intern, I work on the first part, which is to use machine learning and deep learning to forecast what will happen, how many orders there will be for a particular company or particular land on a particular day. My forecast will be used by another group in our company to do some numerical optimization. Also, another thing that I have just done recently was to… When you launch a product, you want to know how good it is. My responsibility was to see if that’s a causal impact of our product on our… that we impact them positively.
Jon Krohn: 00:27:16
Right. To see if there’s a causal impact of working with Ocelot Consulting. What did you find? Hopefully a positive impact.
Khuyen Tran: 00:27:26
That’s a mix, right? I mean, for some region, we can see a statistically significant increases in revenues. But for some region, we see the increase, but it’s not significant, so we cannot do any conclusion on that, but overall increase, so that’s a good news.
Jon Krohn: 00:27:51
Nice. Yes, that is good news. It could just be a sample size issue. It could be that even in those regions where you see an increase, but it’s not statistically significant, it just need more data.
Khuyen Tran: 00:28:02
The sample size is pretty small. We just try to do like, “Oh, we have the data. Let’s see if we can see a causal impact here,” but we also know that the data is not big, so maybe the result can be not accurate.
Jon Krohn: 00:28:20
Right. Right. Right. Exactly. Cool. Sounds like Ocelot Consulting is doing some cool work using data science practically to forecast orders and then to allow freight firms to maximize their profits based on those forecasts. That’s super cool. I have a clear sense of what you do there. Are there any particular software languages or tools or approaches that you use most days in that job?
Khuyen Tran: 00:28:50
Yeah. I write in Python.
Jon Krohn: 00:28:56
Python, yeah.
Khuyen Tran: 00:28:56
Everything I write is in Python. We use Azure ML. There’s a very cool open source software that I use to build data science pipeline called Kedro. I don’t know if you have heard.
Jon Krohn: 00:29:13
How do you spell that? No, I’m not sure I have.
Khuyen Tran: 00:29:15
It’s K-E-D-R-O.
Jon Krohn: 00:29:20
Kedro. Cool. What does that do?
Khuyen Tran: 00:29:23
Right. Writing code in data science is different from writing code for a software, right? The difference is that you have a function, and you use a function to process the data. Maybe you guess how the output would be, but you’re not 100% sure. That is only one function. If you put a blog multiple functions to one data, the code will… It will look very messy, and you even more are not sure about the output. The ideal way to go about doing this is you put your code, your functions into a pipeline, and you know for sure what is the input of the data and what is the output of the data. That’s what Kedro allow you to do. It put functions at nodes, so you have different nodes, and then these nodes are connected through a pipeline.
Jon Krohn: 00:30:28
Oh, cool. It allows you to visualize your whole data processing pipeline so you can keep track of everything that’s going on. That makes a lot of sense. That’s super cool. I’m glad to learn about Kedro today. You mentioned also Azure ML, so Microsoft’s cloud service and their machine learning functionality built into that, and somewhat unsurprising that you’re using Python. That’s cool.
Khuyen Tran: 00:31:02
I just want to briefly mention another tool that is Great Expectations. I don’t know. Is that-
Jon Krohn: 00:31:11
Expectations?
Khuyen Tran: 00:31:11
Great Expectations.
Jon Krohn: 00:31:13
Great Expectations, oh, that’s a great name.
Khuyen Tran: 00:31:20
You have a certain expectation for your data, right? What could happen is your data changed? Maybe it changed in different ways, right? Maybe the column that used to be there for the new data is not there anymore, or there’s missing values that were not missing values before, or there’s a shift in the distribution of your data. Great Expectations can help you to validate your data before you actually apply the functions on it, which is very ideal, because if you have some error, you don’t know if that’s because that’s the new data or that’s your code.
Jon Krohn: 00:32:00
Right. That’s cool. I love these tools. It’s great to learn about these. Thank you so much for sharing those with us, Khuyen. Something that I think might be mind blowing for listeners about you is that despite say doing data science internships at places like Ocelot Consulting, really digging in to juicy, beneficial work really contributing to the company, on top of that and on top of being this prolific writer, both of long-form blog posts at Towards Data Science and more recently NVIDIA as well as on LinkedIn, on top of all of that, you’re a full-time student. You are an undergraduate student. You’re finishing up soon. Next year, you’ll finish your undergrad degree. You’re studying for a Bachelor of Science in Computational and Applied Mathematics, and you’re doing okay at it. You’ve got a perfect 4.0 GPA. How do you prioritize all of these different strands of your life against each other? You’ve got so many different things going on, and you do all of them to perfection. You have a perfect GPA, and you haven’t been writing these long-form blog posts for very long. It’s been less than two years that you’ve been doing it for Towards Data Science, and already, you’re getting hundreds of thousands of views per month. How do you prioritize your time so that you can execute on all of these different strands of your life so effectively?
Khuyen Tran: 00:33:36
Thank you for the very great introduction on that. When we talk about prioritize, the first thing that we want to do is to say no to the things that we don’t care about or not important in our life. For me, I say no to… Except LinkedIn, I don’t do so much social media or… Also, one thing about… One person can sit on a homework for three hours, and that person might not be able to finish as much than another person who work on the homework for one hour, but they very focus on it. For me, I-
Jon Krohn: 00:34:25
Totally.
Khuyen Tran: 00:34:26
For me, I enjoy studying, and I enjoy working, and I enjoy writing. Every time I do something, I pay full attention to the thing that I do. In a way, I’m in the flow. I’m not getting much distracted. That really helps because that pushed me over being tired, so that very helps. I also try not to be stressed, because when you are stressed, “Oh, there are so many things to do. Oh, what can I do?” I think about in the long-term, it would be okay. I will able to finish all these things in the long term. The things I’m anxious about, they will not matter anymore. Because of that, I feel calmer. Because I feel calmer, I don’t bring that stress into my work, which could affect my quality of work in a negative way.
Jon Krohn: 00:35:29
That is such great advice. I suspect, Khuyen, that something that helps with both of those last two things you mentioned… You mentioned three things, saying no, which is hugely important and so easy to not do. It’s so easy to… I’m guilty of this. I’m getting a little better. I’ve historically been very guilty about saying yes to everything, because I like to please people. But if you say yes to everyone, especially when it’s individual requests, that eats into time that you could be spending creating content that gets viewed, in your case, by hundreds of thousands of people. So saying no to some individual requests can lead to a much bigger impact, much, much, much orders of magnitude, bigger impact. That’s hugely important. I agree. For those last two… Saying no was the first thing. The second thing you mentioned was focusing 100% on the task on hand, and the third thing was minimizing stress. I suspect that something that helps with two and three with the focus, as well as reducing stress, is being really passionate about what you do, and you mentioned it there. You said, “I like to study. I like to do my data science work. I like to write,” and so probably that’s a big part of this is finding what you enjoy doing.
Khuyen Tran: 00:36:54
Yeah. Also, there’s a quote called you fake it until you make it, right? I believe that. For the subjects that I don’t like about, I will tell everybody, “I love this class,” even though I didn’t like it. I told everybody, and it was like, “What?” I’m like, “Let’s show.’ I just trick people and trick myself into liking the subject, and because I like the subject, I don’t mind spending hours on doing it.
Jon Krohn: 00:37:27
Right. I like that.
Khuyen Tran: 00:37:27
I actually like them now just for the exception of some classes, but I still like studying.
Jon Krohn: 00:37:37
That’s funny. I haven’t thought of fake it till you make it in that kind of context. I usually think about… In fact, I’ve only ever before thought about this idea of fake it till you make it in the context of just tricking other people into thinking that you know what you’re doing, but I love this way that you’re framing it of even to yourself by taking a class that maybe you’re not so sure about, and just saying, “You know what, I’m going to like this. I’m going to make this work.” Then your own mind comes around and says, “You know what? This isn’t so bad.” I like that.
Khuyen Tran: 00:38:09
Right. I also fake that I’m a good student. If I don’t understand something, I will… If a professor asks questions, and they talk about something I don’t understand, I would just ask some questions, or I just guess what you’re going to say next. People thought I know, but it just about I have more courage than other people to speak up.
Jon Krohn: 00:38:35
That makes so much sense, Khuyen. Tell me, this degree that you’re pursuing, computational and applied math, I think I know what the answer is going to be, but do you think that that degree is going to be helpful to your data science career?
Khuyen Tran: 00:38:54
I would say yes and no. It’s both.
Jon Krohn: 00:38:54
Oh, I’m surprised that there’s any no. To me, it seems like one of the most in demand aspects of the data science career. I can’t… I’m looking forward to hearing the yes, but I’m even more looking forward to hearing the no. Go ahead.
Khuyen Tran: 00:39:08
Right. Right. I guess I… What I observed over time was in my school, I did took some course on math for machine learning, and it was very helpful. I like linear Algebra and all those fun stuff, but what I realized is that if I don’t apply them right away, I will forget them. I think at school, the way they teach you is they teach you a bunch of theoretical thing one after another and after another. Then they… For my computer science class, I had them on different things. Some of them in R and some of them in C++, none of them in Python. [inaudible 00:39:59] like so many software, but the company that you will work with, they do not necessarily use those tools. Once you graduate and you work in the company, you forget about those. That’s why I say that’s a yes and no in there. A lot of knowledge I still remember like the essential concepts of statistics, or linear regression or the time series. I still remember some of them, but some of them I forgot. That’s why it’s nice to.. Another thing that I often advise people on when they want to learn the math behind data science is to do both at the same time. It’s not like you shouldn’t learn the theory, but you should ask a question. Why do you want to learn this, and what do you want to apply it for, and learn a little bit at time. You learn, you apply, and then you learn, you apply one piece at a time.
Jon Krohn: 00:41:01
Getting the theory and the application at the same time can be hugely beneficial. Someone named Vince Petaccio who used to be a data scientist that worked for me, and now he works at Amazon Web Services, he was on an episode earlier this year of the SuperDataScience show. He was on episode number 459 talking about how you can use data science to tackle climate change. Vince, before he worked with me as a data scientist, he simultaneously pursued a master’s in computer science with a data science specialization, which was teaching him the theory. While in his spare time around doing that degree, he was taking Udacity courses, which were hands-on and applied applications of data science, so kind of like you’re saying, or exactly like you’re saying, having these underlying theoretical foundations and making those strong while simultaneously finding data science applications. I think that it’s a great way to make the most of both. Hopefully that also helps with remembering the theory as well. You’re talking about sometimes you learn theory. I guess that’s what your no was, right, that you were saying yes and no. The big no is that you feel like you learn a bunch of things that you just forget anyway.
Khuyen Tran: 00:42:22
Right. Right. That’s how I feel about it. I feel like there’s a better way on doing it. I just feel like if 80% of what you learn is forgotten, then there might be a better way to going about retain your knowledge or the hours that you put in.
Jon Krohn: 00:42:45
Right. I understand. Sometimes it does feel like that in school where you’re like, “I don’t know how this is going to help me.” I can understand that. I can understand, but something that you mentioned to me before we started recording was how you felt that your college degree has helped enormously with learning how to learn. Even if you’re forgetting things later, some of that stuff you’re forgetting is still helping you come up with better ways of learning how to learn, right?
Khuyen Tran: 00:43:14
Yeah, definitely. I think I refine the way that I learn over time. There’s a book called Made to Stick. I read the book.
Jon Krohn: 00:43:27
Made to Stick.
Khuyen Tran: 00:43:28
Yeah. I think it’s either Made to Stick or Make to Stick, but what this one teach you is if you learn a knowledge, and you… There’s two ways of learning, passive and active learning.
Jon Krohn: 00:43:48
Active.
Khuyen Tran: 00:43:49
Passive mean you watch a video, and you’re like, “Hmm,” understand that that’s easy. Active learning mean you watch a video, and you’re like, “Okay, I can’t understand that, but can I actually… What is the application?” You think more of that. You’re like, “What can I do with this, and can I rewrite what he said in my paper, in my own words?”
Jon Krohn: 00:44:17
Exactly, in your own words.
Khuyen Tran: 00:44:18
Right. I think when I read the book and it say like, “The student who study for a long…” They try to role study again and again and again and again. You know how student often stay up late for the test the next day. They just try to read again and again to make sure that they understand the concept compared to the student who take less amount of time, but they actually make themselves feel uncomfortable. When you learn, sometimes I feel easy. Sometimes I feel uncomfortable, but what I learned from the book is feeling uncomfortable when you are learning is a very good sign that you are absorbing new knowledge because your neurons, I think, is something go about your neurons. When you try to learn new skills, when you actually try to… It’s like lifting weight, right? I know you lift a lot of weight, but if you just do the weight that you feel comfortable with, it doesn’t improve much. But if you’re like, “Okay today, I’m going to do a little bit heavier. Today, I’m going to lift heavier,” then you feel it’s hard. Then it is working. That’s how I feel about studying.
Jon Krohn: 00:45:47
That is a great analogy, and I couldn’t agree with you more. I also was one of these people in my undergrad. I was very… I really loved my courses, and I was focused on getting really good grades. Like you, I discovered these kinds of active learning techniques. Just to summarize some of your ideas there, being able to put what you learned in your own words, and thinking about… maybe even actually applying the knowledge that you’ve learned is critical to this active learning being effective. Yes, that analogy, I love that weight lifting analogy, because when you are learning effectively, it isn’t comfortable. If you are finding things very easy to understand, you’re not pushing how much you could be learning in that time. Just like when you go to the gym, if you are not increasing the weights and doing as much as you can, you are making the most of that time working out. Great analogy. In this episode, Khuyen, before we wrap up, we should handle a couple of audience questions.
Jon Krohn: 00:46:59
Sometimes before I have a guest come on the show, I make a post saying, “Hey, this guest is coming up soon. Do you have any questions for them?” I did one for you, Khuyen just earlier this week at the time of recording. We got a huge amount of engagement, which doesn’t surprise me given how engaged your audience is, so over 10,000 views of this post, and so many questions. We had nine questions come up. We’re only going to take time to answer a couple of the questions that you thought were most interesting to you. Nikolay Kurbatov who frequently features on the SuperDataScience podcast with his brilliant questions that he asks guests, he asked five very technical questions. One of them, number two, was what would be your takeaways if you consider the traditional software development approach versus the data science development approach? I think this is an interesting one because you actually addressed this earlier. Earlier on in the episode, you talked about Kedro, which allows you to visualize data science workflows. When you did that, you mentioned how that data science development wanting to see those data flows is different-
Khuyen Tran: 00:48:20
That’s all I got to say about it.
Jon Krohn: 00:48:20
… from traditional software. I don’t know if you have anything else to add, but we’ve covered it. Then so Krzysztof Ograbek who is also big, at least, in my LinkedIn community, he’s very frequently commenting on posts and engaging with people, not just me, but other people that respond to posts that I make. Krzysztof had four insightful, in some senses, philosophical questions. Some of them you’ve already answered. One of his questions was how important is it to show up every day? We talked about that a lot in this episode, talking about consistency. One of his other questions was what’s the biggest difference between today’s Khuyen versus Khuyen writing her first Medium article? I think this is a really interesting question to give you a chance to talk about your journey a little bit.
Khuyen Tran: 00:49:18
For sure. I guess what I did before writing was I tried to be consistent with learning data science by putting one hour per day on doing Kedro computation. How I feel from it is I don’t feel engaged in practicing data science because what I did was I just doing some analysis, but I don’t know where this will go. I don’t have a goal for the project, so it’s not exciting. I thought I didn’t like data science. I thought data science is not my thing, but then I try writing articles, and I try with couple posts that I have a question, and I want answers. I do a data analysis on it, and write an article about it. It was so fun, and I realized that, “Hey, data science actually is fun if I have a goal in my mind on what I want to achieve with it.” Since then… Before, I was not at all confident about my Python skill or data science skill. I feel like I’m a beginner. But once I start writing about what I learn, and people start to recognize me as an expert in the field, I feel much more confident about my skills. I actually increased my skills significantly by writing articles and writing posts every day on Python. I would say me thinking about my… Me before writing, I would think myself as a beginner, but now, I feel much more confident about myself, and I also more confident to tackle more complicated concept instead of say, “I don’t know anything about this, and sorry, but I cannot do this.”
Jon Krohn: 00:51:23
Nice. That’s a great answer. Some things have stayed the same from Khuyen when she wrote her first Medium article versus the ones that she’s writing today. One of those is the consistency. The consistency has stayed the same, but you feel much more confident now than in the beginning. That’s a big difference. Very cool Khuyen, thank you for taking the time to answer those audience questions. Well, a question that I always ask guests on the show, and you already gave us a book recommendation just there, that book Made to Stick, but do you have any other book recommendations for us? It sounds like you read a lot, so you might have more than one.
Khuyen Tran: 00:52:05
I read… I love book on paper form and on audio book, but there’s a book that… I would talk about a book that motivated me to start writing on LinkedIn, because I write article before LinkedIn post. I didn’t write LinkedIn post until I read that book. It’s called Show Your Work by Austin-
Jon Krohn: 00:52:37
Show Your Work.
Khuyen Tran: 00:52:38
… Austin Kleon, I think. I don’t know how to say his last name. What it say about is whatever you do, no matter how crappy your work is, just show it to people, because some people will benefit from it. I was like… You will get better over time. No matter how bad it is, just show your work. At the beginning, I was like, “Okay, this is something very stupid.” I think like everybody know about this is some Python snippet. I think everybody know about it, but I’m going to post it anyway. I don’t mind if people think I’m stupid, but I was actually care a lot about it. I try not to check my notification. People react very well to it. I was like, “Okay, I’m going to just keep doing this every day. So even on the day that I don’t feel like it, I’m going to do it very, very small code snippet, and I’m going to pose it.” What surprises me is something that I talk very simple to people, something I thought, “There’s no way people don’t know about this.” A lot of people don’t know about that. I think we all have something worthy of sharing.
Jon Krohn: 00:53:55
There’s a number of points there that I think are brilliant. I couldn’t agree more that sharing your work is key. It takes courage initially to think, “Oh, if I write a Medium post, or I write a LinkedIn post that anybody’s going to care,” but it helps you refine your own thinking about a concept. So even if literally nobody reads it, that act of listening that you were describing, putting something into your own words and publishing that, something that you wanted to learn anyway, it means that you’re more likely to think critically about this concept and where are the gaps in my understanding. So even if nobody looked at it, it would still be a valuable exercise. But on top of that, as you point out, it’s interesting how often even the stuff that you don’t think is going to be interesting is there are people out there who are more beginner than you. In fact, in a field like data science or software development, there’s way more people out there that know nothing about it or just a tiny little bit who would love to have somebody explain even relatively simple concepts in an easy to understand way, so hugely valuable. This ties into something that I’ve ascribed to for years. I think you do as well. You talked in the beginning of this episode about consistency. You talked about with Towards Data Science, you committed to writing short articles but on a consistent schedule. I think this is the key thing as well.
Khuyen Tran: 00:55:32
That leads to another book that… It reminds me of another book that also really helps me along the way is called Atomic Habit.
Jon Krohn: 00:55:41
Yes, that’s exactly right. That’s where I get these ideas from. I’ve been… Long before James Clear wrote that book, Atomic Habits, I met him on a bus in Switzerland.
Khuyen Tran: 00:55:54
Wow.
Jon Krohn: 00:55:54
He’s sitting next to me on the bus.
Khuyen Tran: 00:55:57
Before he wrote the book?
Jon Krohn: 00:55:57
He said, “Oh…” This was nine years ago, something like that.
Khuyen Tran: 00:56:02
Wow.
Jon Krohn: 00:56:04
It was… No, not quite that long ago. It was 2014. I met him in a bus in Switzerland. Is that six years ago, seven years ago? It’s hard to do math while I talk at the same time. I think it’s seven years ago. He was… At that time, he said, “I am committed to writing a blog post every single Monday and every single Thursday.” He was able to accumulate over half a million subscribers to his blog. They were subscribers to his email newsletter. That’s what allowed him to have lots of content for his book, Atomic Habits, and that it also meant that he had this large audience. Through his posts, I was learning these kinds of ideas like writing consistently on a schedule. I use that as a model for my own life. With all of my LinkedIn posting and any of the writing that I do, it’s on a schedule because it’s crazy. If you publish stuff on a schedule, you force yourself to do it. Posts that you think are short or quick, you didn’t really think critically about it, you didn’t do a good job, some of those posts do really well.
Khuyen Tran: 00:57:16
That’s more like… I think just at the beginning, right, when you try to start something, don’t focus too much on the quality. Try to focus on the quantity, which sounds bad, right?
Jon Krohn: 00:57:29
Exactly.
Khuyen Tran: 00:57:30
I mean, for my articles, I’m like, “If you want to start an article, just write any kind of articles.” Just start writing very short article. If you want to have one hour, write a one-hour blog post, and you end with your day. I think just put your work out there, and have a schedule, right? For me, my schedule is write at least one article per week. I used to have more strict articles. I need to write every Wednesday, and publish on Tuesday, but I realized that sometime I want more flexible than that. Sometimes I prefer to write very long articles. Have some schedule in your mind, definitely will get you through the momentum to get started.
Jon Krohn: 00:58:23
Exactly. Couldn’t have said it better myself. Wonderful, Khuyen. Well, you’ve been an amazing guest on the show. We’ve learned so much about how to be a great technical writer as well as how that technical writing can be impactful for your career, whether it’s data science or something else. Thank you for sharing your secrets with us, and hopefully we’ll have you back on for another episode soon.
Khuyen Tran: 00:58:49
That would be great. Thank you for having me on your show. It was very enjoyable to talk to you.
Jon Krohn: 00:59:02
Khuyen is so incredibly impressive. I’m delighted that she was able to prioritize an appearance on SuperDataScience amongst all of the pursuits she’s tackling with aplomb in her life. In the episode, Khuyen filled us in on how you can maximize the impact of your technical writing by running your content by a non-expert before publishing, by finding ways to explain concepts simply, and by including lots of visuals. She talked about how data science can be used to forecast freight orders, and optimize transport firms’ profitability, the software application, Kedro, for visualizing data flows, the great expectations tool for documenting data science and ensuring data quality, and how to be as epically productive as her by saying no, focusing 100% on the task at hand, enjoying what you do, framing your obligations in a positive mindset and actively learning by applying your knowledge and putting it into your own words.
Jon Krohn: 01:00:00
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 Khuyen’s LinkedIn and website as well as my own social media profiles at www.superdatascience.com/521. That’s www.superdatascience.com/521. If you enjoyed this episode, I’d greatly appreciate it if you left a review on your favorite podcasting app or on the SuperDataScience YouTube channel. I also encourage you to let me know your thoughts on this episode directly by adding me on LinkedIn or Twitter, and then tagging me in a post about it. Your feedback is invaluable for helping us shape future episodes of the show. Since this is a free podcast, if you’re looking for a free way to help me out, I’d be very grateful if you left a rating of my book, Deep Learning Illustrated, on Amazon or Goodreads, gave some videos on my personal Jon Krohn YouTube channel, a thumbs up, or subscribe to my free, spam-free and content-rich newsletter on jonkrohn.com.
Jon Krohn: 01:00:58
To support the SuperDataScience company that kindly funds the management editing and production of this podcast, you could consider creating a free login to their learning platform at www.superdatascience.com, or consider buying a usually pretty darn cheap Udemy course published by Ligency, a SuperDataScience affiliate such as my own mathematical foundations of machine learning course, which I recently finished. You can check out all the linear algebra and calculus content that I’ve published into that course.
Jon Krohn: 01:01:29
All right, thanks to, Ivana, Jaime, Mario and JP on the SuperDataScience team for managing and producing another fun episode for us today. Keep on rocking it out there, folks. I’m looking forward to enjoying another round of the SuperDataScience Podcast with you very soon.
