SDS 073: How to Stand Out to Recruiters in Data Science - SuperDataScience - Big Data | Analytics Careers | Mentors | Success

SDS 073: How to Stand Out to Recruiters in Data Science

Welcome to episode #73 of the Super Data Science Podcast. Here we go!

Today's guest is Founder of Break Into Data Science, Mark Meloon

Subscribe on iTunes, Stitcher Radio or TuneIn

Looking to move into (or within) the space of data science? This is going to be a value-packed episode full of top tips for you as Mark Meloon brings his years of expertise in the field, getting back into the field, and helping others to do the same.

Aside from being a highly experienced data scientist, Mark is active in the community, answering career questions and offering advice on how to get interviews and then excel in them.

As an added technical bonus, Mark will also share an overview of Bayesian Network Analysis and how it differs from Naive Bayes.

Tune in now to learn the secrets of how to have the jobs come to you!

In this episode you will learn:

  • Why Answer Questions on Quora? (6:33)
  • Benefits of a Polished LinkedIn Profile (11:56)
  • Getting into Data Science Without a PhD (16:58)
  • Walk-through of a Data Science Interview (21:58)
  • Is Machine Learning Data Science? (24:30)
  • Communication and Presentation Skills (26:36)
  • Overview: Bayesian Network Analysis (32:32)
  • How To Turbo Boost Your LinkedIn Profile (37:52)

Items mentioned in this podcast:

Follow Mark

Episode Transcript

0

Full Podcast Transcript

Expand to view full transcript

Kirill: This is episode number 73 with Founder of Break Into Data Science, Mark Meloon.

(background music plays)

Welcome to the SuperDataScience podcast. My name is Kirill Eremenko, data science coach and lifestyle entrepreneur. And each week we bring you inspiring people and ideas to help you build your successful career in data science. Thanks for being here today and now let’s make the complex simple.

(background music plays)

Welcome back to the SuperDataScience podcast. Super excited to have you on board, and today we've got a special guest. Mark Meloon is the Founder of Break Into Data Science. And what you need to know about Mark is that this is a person who is extremely passionate about giving back to the community. So if you Google questions about data science careers, very likely you will find answers on Quora.com, and very likely those answers will be by Mark Meloon.

In fact, that is exactly how I came across Mark. I was looking for different opinions on data science careers, and one of the answers was on Quora and was by Mark, and when I looked at his profile, I found that he answers lots and lots of questions relating to data science careers, and how to get the careers, and what the different careers entail, and what a career in data science generally is. And so without any shadow of a doubt, I knew that this is a person that could bring a lot of value to the podcast, and that's exactly what happened.

In this episode, we talked about Mark's personal journey in terms of how he got into data science, and what specifically inspired him to start blogging about data science careers, start helping others, start answering questions, create breakintodatascience.com, and other sources and resources. And also of course we talked about data science careers and solving the questions that Mark has gotten on Quora, some of the more interesting ones, and finally, Mark is an expert on helping people set up their LinkedIn profiles, so was kind enough to share some tips on setting up your LinkedIn profile in a way such that you don't have to look for opportunities any more, that the opportunities will come to you themselves.
So there we go, that's what you will learn from this podcast, and of course, much, more more. And without further ado, I bring to you Mark Meloon, the creator of Break Into Data Science.

(background music plays)

Welcome everybody to the SuperDataScience podcast. Today I've got a very special guest, Mark Meloon, the Founder of breakintodatascience.com. Mark, welcome to the show. Thank you so much for taking the time to be here.

Mark: Well thank you for having me.

Kirill: So, Mark, it was so exciting to hear back from you, because the first time I messaged you was on LinkedIn after I read some of your answers on Quora, where you were helping people on how to become data scientists. That was so exciting. Tell us a bit about yourself, like where are you from, and what do you do?

Mark: Sure. I am a data scientist at ServiceNow, which is a leading Software as a Service company, and we make an enterprise-grade workflow management product. And we've got a number of different offices, but I'm based in the Silicon Valley headquarters. And at ServiceNow, I work in the Business Intelligence group, and specifically focus on improving our sales and marketing processes. So stuff like churn reduction, and lead scoring, and so forth. I've only been with them 6 months so far, but it's really turned out to be a dream job. And like you just mentioned, in addition to my day job, I field questions about job hunting for data scientists on Quora, LinkedIn, Facebook, and, in the next couple of months, I'll be using my blog, which is simply markmeloon.com, as the central hub for all my data science content.

I've got a blog there right now, but I'm going to be posting a lot more information there and I'll still be active on Quora and LinkedIn and so forth, but what I'm going to do is link my popular posts from my main site, my blog, so that if I post something on Quora, I'll post a little blurb on my blog and say "hey, I wrote this answer on Quora, it's gotten a lot of upvotes, why don't check it." So that way, people don't have to try and follow me everywhere.

Kirill: Fantastic. And I'm very excited you mentioned Quora, because your answers on Quora and how you field these questions, it's just fascinating to see. I was very impressed by your answers. And just for the benefit of our listeners, so you guys get a sense of what kind of questions we're talking about, and what Mark's expertise is in, I’m just going to read some of the titles of the questions that Mark answered on Quora. For instance, “Do you need a PhD to become a data scientist?” “What kind of career progression can one come to expect being a data scientist?” “What work does a data scientist do 80% of his or her time? To be more specific, what is the most important aspect of a data scientist?” “Does it make sense for a data scientist to learn Python after gaining proficiency in R? Is it not better to learn Java for production purposes?” “I have six months to learn data science. Should I study as many courses as I can or just get onto Kaggle and practice?” and many, many more.
I’m just going through a list here on Quora. There’s like dozens of questions that Mark has answered and that is very, very exciting because it’s very helpful to people. Even I found some very valuable information on how people can approach data science. And I hope you don’t mind, Mark, I’ve borrowed some of your ideas to share with students or people who have asked me how to get into data science. And I thought it’s just going to be easier to get you onto the podcast so you can share all the vast experience and knowledge that you have so people have this one place where I can tell people to go to listen to you speak. Tell us a bit about why. The first question that comes to mind is why do you go out of your way to help people—to answer these questions for people and help people get into data science and find the right career paths for themselves?

Mark: Well, there’s really two main reasons why. Number one is, there is a bias in hiring for data scientists that have a Master’s or a PhD. I don’t think that’s really the best way to go. I think people with a Bachelor’s and with passion, and especially if they have some kind of experience, they make very good data scientists. You don’t really need PhD level mathematics or something like that in order to do good data science. So because there’s this bias, I want to counteract that and give beginning data scientists an edge in the job seeking process so they stand out from the other applicants.

But the other reason is that I had to figure out how to be hired as a data scientist myself. And very briefly, when I came out of school I got a job with a small defence contractor. I was there for 13 years, it was a great job. That was where I actually experienced data science, because in my schooling, I didn’t do anything with that at all. I worked there for 13 years and then I decided to go off and start my own company. I had done that for 2 years. Believe it or not, it was social media marketing and I wanted to bring the power of data science to the small business. But I found that working by myself didn’t really suit me, so after 2 years I shut the business down and went back into the corporate world.

And I’ll never forget that I had a conversation with a recruiter who specialized in placing analytics jobs. So I sent her my resume, and I was on the phone with her, and I said, “Okay, well, now that you know a little bit about me, can you recommend me to some employers and connect me?” She said, “No,” and I thought “What the heck?” I was a star performer at my previous job and I asked her why. And she said, “I understand that you did some data science in your earlier career, but your job title was Senior Analyst, not Data Science, and then you took two years off here to do your own thing, and now you’re trying to come back into the hottest field. I think your skills are probably out-of-date and quite frankly, I think your un-hireable. I think you’re a mid-career scientist who is trying to break into something that they don’t necessarily know anything about. And quite frankly, I couldn’t make any money off you.”

You know, I was shocked, but I realized that I got that first job right out of school. This was in the late 90s and since that time, the Internet had really changed how one looks for a job. You know, coming out of school, companies seek you out rather than the opposite. So I realized, “Wow, I don’t really have any job seeking skills.” So I had to develop those, and especially I had to come up with a good story for what I had been doing those last couple of years and why I’m still a relevant data scientist.
So a lot of effort, a lot of research on how to get a job, especially as a data scientist, and I don’t really want other people to have to go through that because it took a lot of effort. So now that I have developed a system for myself, I’m eager to share it with other people because it made a huge difference for me.

Kirill: That’s a fantastic story. And there’s so much in it. I hope we’ll go through a lot of those elements in the podcast. But probably the first thing that popped into my mind when you were describing this was that having your own business, you realized that that wasn’t for you. And that’s a very big thing for you to say and to admit to yourself. And that’s a very important thing as well, because I find that a lot of people are chasing this idea that they have to have their own business, they have to work for themselves. And ultimately, it’s a very different type of work, different type of lifestyle, different type of problems that you have to solve. And that’s what I also keep saying to students, that sometimes maybe you are going to be much more happier being the best, a really good data scientist, being the best in that specific area that you’re in, and progressing your career there, rather than going out there and trying to start your own business.

So I really appreciate that example and you sharing that, that you started your own business, you gave it a go, you put everything into it, and then you thought, “Oh, actually no, I’m happier doing what I’m passionate about, which is data science.” That was really powerful for me. Now, moving back to the issue that was there, that the recruiter was kind of reluctant to help you out, that’s also an interesting case. Even though you know you have the skills and you know you can do the job, the in-between person that makes these introductions, that helps you find jobs, they’re not confident about that and they have a different mindset about that. So you go off and you create these opportunities for yourself.

So, LinkedIn is probably one of the most powerful tools right now to use. Would you say that aspiring data scientists have to have LinkedIn? Just to clarify for everybody listening on this podcast, if you don’t have a LinkedIn – Mark, what do you say to people who don’t have a LinkedIn right now?

Mark: Well, they’ve definitely got to do that. And they really need to put some serious effort into making their LinkedIn profile really good. And I can tell you that in my last few job searches I have not reached out to anyone. They have all come to me, you know, whether it’s hiring managers, or whether it’s recruiters. And quite frankly, I’ve just been inundated with people reaching out to me. But that’s because I put a lot of effort into my LinkedIn profile.

First of all, making it so that recruiters who are doing a search can find me, but then also make a compelling case that will have them reach out to me and ask for my resume or ask for a phone call or something like that. So absolutely, you have to have a profile and put some effort into it, too. I have a lot of people reaching out to me on LinkedIn and Facebook and so forth and they ask me if we have a job or we’re hiring for interns or full-time people at ServiceNow, where I work. And I take a look at their LinkedIn profile, you know, and a lot of them are really terrible. I had one guy I think just last week who reached out to me and put a link to an opening – we do have them at ServiceNow – and said, “Can you refer me to this position?”

First of all, I don’t know the guy, so I wouldn’t have referred him anyhow. But secondly, I took a look at the job description and it was for a Senior Data Scientist. So I looked at his profile, and his tagline was ‘Data Science Intern’ and I said, “Whoa, whoa, whoa, I’m not recommending you from an internship to a Senior Data Scientist.” And he wrote back and said, “Well, I have three years’ experience in the field.” So I took another look at his LinkedIn profile and yeah, sure enough, before his internship, he does have 3 years’ experience, but I completely missed that the first time around just because he had listed himself as a data science intern.
I’m a believer that, regardless of your experience, you ought to call yourself a data scientist if that’s what you want to be. You may be a data scientist who is currently unemployed, and I see a lot of people who list ‘seeking data science opportunities’ or ‘data science enthusiast’ and I tell people, “No, you’ve got to remove that. It makes you sound like an amateur. It makes you sound desperate.” Like I said, I completely missed that this guy had several years’ experience just because his profile was not very clear and he didn’t put the important stuff upfront.

I know that’s kind of a long answer to your question, but yeah, LinkedIn is critical, because that’s where you’re going to get your opportunities from. And even if people don’t reach out to you and you submit an application somewhere else, the first thing they’re going to do is look at your LinkedIn profile.

Kirill: Exactly. I totally agree with this notion of seeking opportunities. And not just in data science, in anything. Like, I’ve seen titles such as ‘Seasoned Executive Seeking New Opportunities.’ It does come off as a bit desperate and I personally wouldn’t put it there. I would just put the title of the job that you’re aspiring towards. It’s not necessary, if you want to be a data scientist and your current position is a data analyst, or your previous position was data analyst, you can still put ‘Data Scientist’ and it just shows that that’s what you want to be, that’s what you’re aspiring towards, and then slowly add more stuff to your LinkedIn as you do more courses or as you have some more experience. And also it will be a guiding force for you, so you know, “That’s what I want. That’s what I want on my LinkedIn.”

Mark: One thing that people don’t realize is that one of the important things about a LinkedIn profile is coming up with keywords and key phrases that people may be searching for. So a lot of people for their tagline do put down their job title or something like that, but I advise people, “Look, just sprinkle it with a couple of keywords.” When I was looking for my last job, my tagline was ‘Data Scientist │Predictive Analytics │Machine Learning.’ That’s not a job title, but I put in three different keywords that I knew people would be searching for because it turns out LinkedIn’s algorithm is such that your profile will be ranked highly as long as you have some keywords in a couple of important strategic sections, and that tagline is one of them.

Kirill: I totally agree with that and that’s very powerful. I would love to get some of your LinkedIn tips. We’ll get back to that in a bit. For now, let’s just go through some of the most common questions that you’ve had on Quora or other places that I’m kind of interested in. I probably didn’t give you a heads-up about this, but I’m sure it should be fine. Are you comfortable if we jump through a few of your questions on Quora?

Mark: Sure.

Kirill: Okay. One that really stands out is, “Do you need a PhD to become a data scientist?” You’ve already mentioned that the answer is no, but could you elaborate a bit more on that and give a bit of comfort to the people listening who are always worried or always heard that you need a PhD to be a data scientist and things like that?

Mark: Yeah. That’s a great question. In fact, I was on a panel recently at Galvanize in San Francisco, which is one of these data science boot camps and that question came up. The answer is no, you don’t need a PhD to become a data scientist, but as I mentioned earlier, there is a bias towards them. And for a lot of people, a lot of beginning data scientists, they don’t really understand why.

And the reason why is because a PhD demonstrates that you have some experience solving problems that aren’t necessarily spoon-fed to you. You know, in a lot of courses there’s projects and there’s homework sets and so forth, but the data is oftentimes cleaned and prepared for you, a lot of times they tell you exactly what algorithm to use and so forth. As you know, that’s not what the real world is like at all.

So, when you get a PhD, you have to do independent research to do your dissertation. So when people see a PhD, they already know that these people have experience working on projects. Similarly with a Master’s, people have to develop a Master’s thesis. Obviously, what you want to do if you have a Bachelor’s is you have to demonstrate that experience as well. You can either do that through previous job experience, professional job experience, or if you don’t have that, if you’re coming right out of school, then doing your own personal projects is critical because that will demonstrate that you have experience acquiring data, doing the data munging process, trying to figure out how to take a general problem and cast it into a data science problem that’s interesting and solvable, and then also communicating your results.

So, these projects that I recommend people do are more than a few weeks in length. They should be a couple of months. It should take you some effort. And the projects that I think are best are these end-to-end projects, where you have to acquire the data and do all these sorts of steps because that will demonstrate that you’ve done real data science problems. Yeah, you don’t need a PhD, but without that you do need some sort of projects, or portfolio, or something that you can demonstrate, “Hey, I can do this, too.” The PhD isn’t really necessary, it’s the experience that’s necessary.

Kirill: Gotcha. So, PhD is kind of a stamp of approval that this person knows how to do research and has experience. You just need to find other ways to do that, to demonstrate that you have that, and you’ll be fine.

Mark: Exactly.

Kirill: I totally agree with that answer. Another interesting one you had on Quora was “What is a senior data scientist?” Like, not just a data scientist, but a senior data scientist?

Mark: Yeah, that’s a great question, because I know that on there, you know, there were a couple of other people who said, “Oh, it’s just a marketing term. There’s no difference.” But my experience has been that it’s considerably different. And to give you an example, I do a lot of data science. I do a lot of machine learning. That’s day-to-day what I do. But another important part of my job is giving presentations, working together with subject matter experts.

So a senior data scientist oftentimes is responsible for—you know, in data science there is this process called CRISP-DM, which talks about starting off with business understanding, turning it into data understanding, then data preparation and modelling and so on and so forth. So, a junior data scientist will probably be given just the machine learning part, doing the modelling, maybe some data cleaning and so forth. But the senior data scientist really needs to be good at the entire scope, the whole end-to-end process.

So, a lot of times, customers or stakeholders or people who you are doing this project for, they have only a vague sense of what they want, so you have to sit down with them and figure out, “Okay, based on this vague idea which they oftentimes haven’t really thought through in great detail, how can I convert that into a data science problem?” So there’s a lot of back and forth. There’s a lot of thinking about, “Okay, what’s doable and what’s important?” And then at the end, there’s also this part about communicating your results in not only a clear manner, but also one that’s compelling so that the stakeholders, the decision makers, will actually take action. For me that’s the really big difference, is that the junior data scientists can focus on some of the details of modelling and data preparation, but the senior data scientist has to be able to do that plus a whole lot more.

Kirill: I totally agree with that. That’s a pretty apt description. I think that’s how it is in consulting firms as well, so that’s valuable. And another one, to finish off, is an interesting one, I think. “What are some data science interview questions and do they include algorithm questions?”

Mark: Oh, yeah. You know, it’s funny, there’s a blog post by Erin Shellman, who is a well-known data scientist online, and she has a great article called ‘Crushing It: How to Prepare for Data Science Interviews,’ or something like that and she starts off her blog posts saying data science is interdisciplinary, which is code for “You need to know everything about everything,” which in some sense is kind of true. You know, you go to some interviews and it’s all about optimization, and then you go to another one and it’s about game theory and so forth.

So there’s a wide, wide range of questions you can be asked in interviews, but what I find is that the majority of interview questions are about how well do you know the basics of popular machine learning algorithms. So it’s not necessary to have a huge breadth of knowledge of algorithms, some esoteric ones, but you really do need to know linear regression, classification and regression trees, Bayesian approaches. You need to know those pretty well that you can know when to use those, what are some of the drawbacks, and certainly be able to explain them.

So those are a lot of interview questions that come up. You know, a lot of times when you are going for site interviews, they’ll have you do whiteboard exercises, and there they’ll give you a problem that may or may not be fully defined, so you will have to figure out how to go about solving that and what questions and assumptions you can make. And hopefully your viewers have encountered this a little bit, because one of the later steps in the interview process is when they give you take-home problem sets. These are typically 3 to 5 hours timed problem sets where they give you data and they give you a problem and they want to see you solve it and write up your answer. And there, what I found is that it’s really critical to have good data munging skills, because you can spend a lot of that 3 or 5 hours just trying to prepare the data, so that’s not necessarily an interview question, but that’s certainly a skill that you really need to have a good handle on during the interview process.

Kirill: Okay, that’s interesting. What would you say is the difference between machine learning and data science? And is there a space for data scientists without machine learning skills?

Mark: Yeah. Actually, that’s a really good question because a lot of people in the industry tend to use those as synonyms for each other. And what I would say is that machine learning is a part of data science. It’s an important part, but it is just a part because, as I say, data science is largely like the CRISP-DM model, where you have to have a business understanding, and you have to be able to cast that in a form of a data science problem. You have to work with subject matter experts. And then, once that’s all done and you do the exploratory data analysis and you acquire the data and all that sort of stuff, then you can turn your attention to the machine learning, which is really focused on a lot of algorithms and trying to figure out, evaluate the performance of your models, and refine those, and try to speed those up. And then after that comes the communication of the results which, again, I don’t feel it’s part of machine learning, but it’s absolutely a part of data science.

So as far as your question, is there a room for data scientists who aren’t that good in machine learning, it depends. You know, for a junior role you need to be really good at machine learning. For a more senior role, I believe it is possible to have less of a detailed knowledge. You certainly need to know the pros and cons of machine learning techniques and when to use them, and key concepts like over-fitting and data leakage and things like that. But if you’re a senior data scientist, or you’re a data-savvy manager, those become less important and some of these other larger, more top level issues become important.
So I would say it behoves everyone to be good at machine learning, but it is possible to be a data scientist without a strong machine learning background. Getting hired as that is another story, because a lot of the questions you have to field during the interview process for a data scientist are going to be machine learning questions.

Kirill: Okay, that’s interesting. I 100% agree with you on that. I agree in terms of the interview that people will be asked sometimes machine learning questions when even they probably don’t have the expertise or are not expecting them. But at the same time I think that there’s still space for people or students or data scientists who are not necessarily into machine learning algorithms and into all of that very sophisticated, highly technical stuff, but who are more creative, who are better at visualization, for example, who are better at storytelling, who are better at taking a problem, dissecting it, or maybe just somebody who’s good at data preparation, not the machine learning part, but data preparation or maybe visualization with Tableau or Power BI. You can still get some really impressive results without having that machine learning aspect to it.
But I do agree with you that you need to look for the right jobs if you are that type of person. And when you’re out on the interview, you need to make sure you convey the value you can bring even though you don’t have these machine learning skills per se, but you can bring value through Tableau, or through Power BI, or through SQL.
For instance, so many jobs when I was back at Deloitte, during my consulting times, didn’t require the machine learning algorithms. For instance, I knew a couple and even before I knew lots more I was able to deliver huge projects or contribute to huge projects just through the power of Tableau, SQL, Excel and a couple of other tools even without going to the machine learning side of things. Would you say that’s a fair comment?

Mark: Yes. Actually, that’s a really good point. It really does depend on the position that you’re interviewing for, but I am certainly a very big proponent on, like you said, visualization and storytelling. Those are really where the real wins are in data science. And I tell people, “You know, you could be brilliant at machine learning, but if you can’t communicate it in an interesting and compelling way that will have decision makers take action, then there’s a big question. Are you really providing value to the organization?”

You know, I encourage people, when they are looking for data science jobs, to develop these other skills and to use them to stand out from the crowd. Just one example: I encourage people, if they are invited to a site interview, that they volunteer to give a presentation because communication is very important. And as long as you’re even passable at giving presentations, the employer is going to be impressed with the fact that you realize that storytelling and communication and visualization is really important. They’ll appreciate your confidence in being willing to demonstrate that. So, absolutely, yeah, I agree with what you said.

And there’s a lot of people coming out of school with machine learning expertise, so how are you going to stand out from the crowd if everyone’s taken the same course as you have? Well, what you describe, the whole casting the problem into a data science problem and then communicating your results, you know, there’s so many libraries and packages out there that do a lot of the machine learning heavy lifting for you – scikit-learn, the R packages on CRAN, now we have TensorFlow, which is expanding its capabilities. So, again, knowing the real nuts and bolts of machine learning is becoming less important than how you actually use those tools and what problems you actually solve.

Kirill: Yeah, those are some great examples. So, when people come out of universities, they have these machine learning skills a lot of the time, but storytelling is important. And I really liked what you said about when you’re invited to an interview, you should volunteer to do a presentation at the interview. I haven’t heard that before. It sounds like a great tip, it’s something that will make you stand out. Can you give us a little bit more detail? How would someone go about that? They get an invitation to an interview and then what do you do? Do you say, “Hey, I actually want to also present while I’m at the interview?” How do you do that?

Mark: Yeah, I would tell them as soon as they make the invite. A lot of times they’ll list off a timeline for, you know, “At 10:00 you’re going to interview this person, at 10:30 you will interview with that person.” And I would actually reply back and say, “That’s great. And I’d love to take 30 minutes and give a presentation on some of the work that I’ve done or maybe a technique that isn’t widely known.” And a lot of times they’ll say, “Oh, that’s a good idea. Okay, yeah, we can work that into the schedule.”
So, I would just plain ask them to do that. And as far as what to give for a presentation, if you’ve done outside projects, that’s a perfect example. Because it’s one thing to actually list a project on your resume or your LinkedIn profile or your personal blog if you have one, but actually talking about it and fielding questions during a presentation, that’s very powerful.
And the second thing I would do during presentation is talk about a non-mainstream data science technique. A lot of times I would either talk about social network analysis or Bayesian networks, which do fall under the heading of data science, but they’re interesting enough that a lot of people don’t know them. And I found that data scientists are really eager to learn about new techniques. So, those things are inherently interesting and also make you stand out from the crowd.

Kirill: That’s really cool. Tell us a bit about Bayesian network analysis. What does that entail?

Mark: Sure. Bayesian networks are basically probabilistic graphical models. And what that means is that you have a joint probability distribution and you’re going to factor that into the variables that are related and are not related. And you can create what they call a directed acyclic graph of the random variables and draw arcs between them to connect the ones that are dependent on each other and the ones that aren’t dependent on each other.

So this is an advantage over naïve Bayes, where you assume everything is independent. Here you are explicitly calling out the dependencies. And how you actually do that, you can build these networks through a combination of learning them via data, but you can also work with a subject matter expert to come up with the structure first and then the conditional probability distributions and stuff like that. You can learn that from data.
I’ve used those quite a bit, especially when I was working in my first job, because it’s great to be able to take a lot of information and fuse it together, a lot of weak pieces of information to actually fuse it together into coming up with strong conclusions about something. Also, you can use a dynamic version of it called dynamic Bayesian networks to actually model Markov decision processes.
The number of problems that you can apply it to is actually pretty broad and yet, a lot of people haven’t really heard about the approach, so that’s a great example of something where you’re teaching a valuable technique and their eyes open and they say, “Wow! I wonder if we could apply this to our problems.”

Kirill: That’s really cool. And just to supplement that example, that’s an example of a very technical, ‘on the cutting edge of what’s being applied in data science’ model. Again, if you are one of those people who are not into machine learning, who are not that technical, you want to be a data scientist but you are good at presentation, you’re good at maybe visualization, SQL – there’s lots of things to be good at – there’s still things that you can showcase.
So I’m going to give two examples. One is of a student I recently met when I was in Florence. Actually, I hope I will get Emanuele on the podcast. He basically just took visualizations we were creating in the Tableau course and then the Tableau advanced course, and he put them onto Tableau Public and that’s it. Then he just brought those visualizations, or sent those visualizations, to the people that he was being interviewed by and they just had a look and they saw that he can solve sophisticated problems in a very visual manner and they were very impressed.
The second example is from my personal experience when I was interviewing for my job at the pension fund in Australia just after my consulting time. And there what I did is I took some of the work that I did in consulting. Of course I desensitized everything. I made it such that it’s just showcased the techniques that I used. And then I actually printed those out—and once again, they were completely desensitized so the numbers, the names, everything was different. But I printed those out and I brought it to the interview. And at the end of the interview, after I answered all their questions, after behavioural questions, theoretical, mathematical questions, questions on tools and such, whatever they had, just before the end, to completely concrete in this opportunity, to cement it in, and make sure they’re 100% impressed and that they are hooked, I pulled this out of my bag and I said, “This is the work I’ve done before. This is what it looks like. These are the things I can create and these are the things I can bring to your organization.”
And from that moment I knew they were hooked. I knew that’s it, that we’re going to have a great partnership together and we’re going to create some amazing things. Adding that extra to the interview process, something they’re not expecting, always makes you stand out. So thank you so much for sharing that, Mark.

Mark: No, you’re absolutely right. Anything that will make you stand out like that, delivering solid business value, that will get their attention because they’ve interviewed a lot of people before and they have all answered the same machine learning questions. But then emphasizing that you are more than that, that you are a data scientist, it definitely is appealing to them.

Kirill: Okay, fantastic. And now I think we’ve discussed quite a lot of interview parts and questions and what to expect career-wise. Now let’s take a step back. Let’s talk about some of your LinkedIn tips. We already started talking about that, but I think this is the right time to go into detail and in-depth on that. You are definitely a LinkedIn guru and people who are doubting that should check out Mark’s profile. If anything, you can just learn a lot of things just by looking at that. Can you share with us some of your top LinkedIn tips that people can take in action already today?

Mark: Yeah. First of all, and this is something that very few people know or do, on LinkedIn, the very first section is your summary section. And I look at most people’s summary and it’s all about them and “I’m good at this and I can do that.” What you really need to understand in the job search process is that these companies have a problem and they need it solved. So, in the summary section, I use a generic formula called ‘Problem-Agitate-Solve.’ So the first thing I do, my first sentence is I repeat back to them, “Do you have this problem?” If I remember correctly, my last job search I wrote something along the lines of “Do you feel like you’re falling behind at the data science revolution?” That will resonate with a lot of people, because they do feel that way, even if they have a data science capability already. So I would start off with that, you know, “Do you have this problem?”

And then the second sentence was a bold “I solved this!” I let you know that I am the solution to your problem. And then I list a couple of reasons why you should believe that. In this case, the Problem and the Agitate were the same thing. So, you start off with “Do you have this problem?” and then you get them to really think about “Oh, this is really a problem that I need to have solved ASAP.” You kind of—the term is ‘twist the knife.’ You let them know about a problem and then you convince them, “Look, you can’t let this pass any longer. You’ve got to take action.”

And then the action, of course, is, “I solved this and here’s the reason why you ought to believe me.” It’s a different approach. You know, most people’s LinkedIn summary is very focused on themselves, whereas if you write it in this manner it’s focused on the company and their problems. Of course, that gets their attention right away, because it’s showing that you understand their problems and that’s really what they’re looking for. They’re looking for a solution for their problems. So that is one tip I would say right off the bat—

Kirill: So just on that one, you talk directly to the recruiter or directly to the person? It’s not just you’re writing a billboard. You’re writing specifically to the person reading as if it’s like a conversation you’re having?

Mark: Yes.

Kirill: That’s powerful. That’s very powerful.

Mark: That’s a great way of putting it, yeah. The second thing, as I mentioned already, coming up with keywords and sprinkling those in your profile in the appropriate places. Now, you want to do this to the right degree. You know, there were people in the past who would just stuff their profile full of keywords to the point where it was almost unintelligible, and that’s not where you want to go. But I encourage people to pick two or three keywords—and again, when I say ‘keywords,’ I mean sort of key phrases like ‘data science.’ That’s two words, but I call that a keyword.

So putting those in your tagline, putting those in your summary section, putting those in the titles of the first two jobs that you’ve held. And we talked about this earlier, where even if your job description or your job title was not data scientist. If you did data science, you should put that in there. One of my past jobs, my title was ‘Senior Analyst/Business Architect.’ What the hell does that mean? What I did was data science, so I think I changed the title to ‘Senior Analyst/Data Scientist.’ Again, I knew that that would help me come up in search rankings when people did a search for data science. So that is another thing. And then the third tip that I would give people is to use a professional headshot, a really good-looking headshot. You know, this seems like something that shouldn’t matter, like maybe it’s a cheap trick or something like that. I’ve put in a professional photo of myself which not only was taken professionally, but I actually had a graphic designer friend of mine actually put a background of a blackboard full of mathematical equations. That was not part of the original shot, he photoshopped that in. And he also cleaned up some of the things on my face; I had some oily sheens so he cleaned that up and so forth so now my photo looks pretty good. And again, that gives me the impression of being very professional.

Again, you wouldn’t think that would matter, but it really does. I see a lot of people’s LinkedIn profiles and they look amateurish, and that then carries over to how I view their profile. “Oh, this is just someone who doesn’t really have much experience and they’re not really a professional yet. They’re still in training. I think I’ll pass on that.”

Those are probably three of the most important LinkedIn tips that I can offer. And I actually have a lot of tips on LinkedIn, because as you say, I’m pretty skilled with it. I was actually interviewed as part of a general job seeking program that someone sold actually for $1,000 and an interview with me via Skype was one of the modules in there. I have a lot of tips that I’ve actually put down into a whopping 17-page guide that is free. I put that available on my website, markmeloon.com where I go through and describe some of the things I’ve already mentioned here, but other tips as well. I encourage every data scientist to download that, especially because it’s free, and start implementing those suggestions because they can make a really big difference.

Kirill: Wow, thanks a lot, Mark. First of all, those three tips are already super valuable – the summary section, then coming up with keywords for the title and using a professional headshot. Also thanks a lot for sharing the 17-page report for free. That’s really cool. So guys listening to this, you can get it at markmeloon.com. I’m just checking it now, here on the right is like a box where you can download that report and probably get some really valuable tips from there.
Okay, so that’s how you structure your LinkedIn profile. Those are just three things, plus there’s 17 pages of other tips that you have. Probably we’ll get people up to speed on that. Now, how do you get people to come to your LinkedIn profile? You said that when you were looking for a job previously, you didn’t have to go and engage recruiters yourself, they came to you. How do you make that happen?

Mark: Yeah, that’s a great question. Part of that is by sprinkling those keywords in the appropriate section because—you know, there’s such a demand now for data scientists that recruiters and hiring managers are doing search using LinkedIn’s search engine. And if you put those keywords in the right area, you will come up high in the rankings.
Another thing is that there is a paid option for LinkedIn – I believe it’s called ‘Job Seeker Package’ – and the last time I looked it was around $30 a month, which is somewhat of a cost, but what it does is it basically affects your chances of getting found. If you have one of these, when someone does a search for you, your profile has a better chance of appearing on page one. That’s just one of the benefits that LinkedIn offers you if you pay a little bit of money. I would say those are two things for helping yourself get found.

Another thing, and this is something that I honestly didn’t do as much as I would now if I knew then what I knew now, is to be active on publishing content on LinkedIn. You know, there are people out there who will take third party articles and post it on the timeline. You know, that’s one approach, but I recommend that if you’re going to do that, and I do recommend that you do that, that you add in a paragraph or so of your own thoughts and interpretations because you want to showcase yourself. You don’t want to necessarily showcase whatever is in this third party article. Granted, you want to pick articles that are interesting, but you want to put your own spin on it and let people know that you do know what’s being talked about and also express your philosophy and demonstrate your knowledge of it. So that’s one thing.

There’s also now the ability to create long form blog posts on LinkedIn through something called LinkedIn Publisher, so I encourage people to do that. And lastly, something that I have done which has worked out really well is, when other people post the articles and things in the timeline, that you respond to them with insightful comments. Again, not like a sentence or two. I know that the character count is 1,000 characters because I’ve written long, long replies to people and then had to cut it back in order to fit it in there, but again, people recognize that “Oh, wow, this person really knows what they’re talking about and they have a lot of insight into this problem.” You know, it’s natural then to take a look at the profile and find out more about this person who wrote all this great stuff.

Kirill: That’s some fantastic advice. Thank you very much for sharing that. Okay, I think those are some very good tips on LinkedIn and how to make sure you’re recognized and that people who do find you, because people will find you, especially if you’re applying for jobs – as you said, Mark, they will come to your profile anyway – to make sure that those people are impressed and they realize that is what you’re looking for.

And just to conclude, would you agree that your LinkedIn profile has to constantly adapt, it has to constantly change, especially while you’re looking for a job, it has to match the type of roles that you’re applying for? If you’re applying for roles in machine learning, specifically folks in machine learning have to have machine learning keywords throughout. If you then change your mind or six months down the track you’re now applying for jobs in the space of data science or analytics or maybe this is your next time you’re applying for jobs, that you’re applying for jobs in the space of data science and visualization, storytelling and SQL and things like that, then you have to redo your whole LinkedIn profile and retailor it towards that. Would you agree with that comment?

Mark: Yeah, I would. That’s a really good point. One thing I encourage people is, “Look through the job listings of those companies that you’re interested in. Notice the words that they use, notice the phrases that they use, and then repeat them back to them in your LinkedIn profile.” So, yeah, I encourage people to do some solid research into the companies and the types of roles that they’re really interested in and learn more about what their phrasing is.
And just as a simple example, you know, do you want to put down text mining as one of your skills or NLP? You know, there is a difference between the two, but a lot of times people use them interchangeably. So you want to look through some job descriptions and see, in the particular type of work that you want to do, which one is more prevalent and then put that in your profile. And yeah, I am a big believer in overhauling your profile.
One of the neat things about LinkedIn is that it shows you, especially in the job seeker profile, it shows you how many people are looking at your profile and who they are. If you just do some basic analytics, maybe just record your weekly number of views in a spreadsheet, you can start to see – it’s kind of like A/B testing. You are going to test your profile and see which variations of your profile are getting not only the most number of views, but the more number of people reaching out to you.
There’s two things that your profile needs to be good at. One is attracting attention, coming up in searches, but then when people find you it has to be compelling enough that they’ll actually reach out to you. There’s two different things there, so trying to get the right balance and trying to figure out the right wording and so forth, that’s going to take a little bit of experimentation.

Kirill: Yeah, fantastic. It’s interesting because you’re doing analytics to get an analytics job, it’s like when you’re watching a movie and they mention the name of the movie in the movie. That type of situation. Yeah, that’s pretty cool. All right, thank you so much for sharing that. I have one question that I’m really interested to get your opinion on before we wrap up. You’ve obviously helped so many people get their LinkedIn profiles up to speed and understand better about data science careers and what they should be looking for, what they can be looking for. From your point of view, from everything you’ve seen, where do you think the field of data science is going and what should our listeners look into to prepare for the future?

Mike: Yeah, that’s a really interesting question. And I’ve actually been doing more thinking about that lately at the job that I’m at because we’re starting to think about, “Okay, how can we build up our team? What is wise for us to make our investments in and so forth?” And honestly, I think there’s going to be less of an emphasis on the details of machine learning algorithms and more on things like exploratory data analysis and understanding/casting business problems into tractable and useful data science problems, knowing how good an answer needs to be in order to satisfy requirements, and of course, all forms of communication.

A lot of these things we’ve already discussed in this interview in various forms here, but I think there is a—I don’t want to say ‘real danger,’ but I think some of these more machine learning aspects are going to become more and more automated. And already there’s many high quality libraries that take care of the details of the algorithms for you, like scikit-learn and the R packages and TensorFlow. And then there are these new platforms that aim to be a so-called ‘data scientist in a box’ that try to empower business analysts by automating some of what we data scientists do.

And I can see a very near future where a data scientist just needs to know the pros and cons of using a gradient boosting machine and maybe just a vague idea of how it works. This may horrify some purists, but I do think that’s the way we’re heading. There’s a school of thought that anything for which there exists a detailed process, whether it’s blue-collar or white-collar work, will be automated in the next decade or so.

So even though modelling is the sexy part of our sexy science, it may very well be the first part that gets automated. I’m sure you and your audience have seen flow charts and cheat sheets floating around the web that tell you what modelling technique to try based on how many variables there are or what kind of features you have and so forth. And I think stuff like that is bound to be automated very soon.

Just as a relevant example, just a few weeks ago, Google claimed they have developed this sort of ‘one model to learn them all,’ some kind of multi-model that can be applied to a variety of tasks. Now, who knows where this avenue of research will go, but if it proves fruitful, the number of useful machine learning algorithms may actually dwindle. So I think data scientists, especially beginning data scientists in particular, ought to start building up the skills that we’ve talked about that will be hard to automate. First, this will make them stand out from other applicants, and then secondly, I think developing these skills will make sure you aren’t automated out of a job in five years’ time.

Kirill: Fantastic. Thank you so much for sharing some very valuable advice there. So, guys, look into skills—even in data science, look into skills that are very unlikely, or less likely to get automated. Those involve things like creativity, storytelling, understanding not necessarily the mathematics behind an algorithm, which is important sometimes, but more importantly understanding the intuition behind an algorithm, when to use which algorithm, you know, pros and cons like Mark mentioned.
Thank you very much, Mark, for sharing that. It’s been a pleasure having you on the podcast, and I’m sure lots of people are going to learn a lot from what you had to share. If our listeners would like to contact you, follow you, or maybe send you some follow-up questions, what is the best way to find you or to follow your career?

Mark: The best way is, as I say, following me on my blog which, if you go look at it right now, is a little sparse and there’s a wide range of topics. But now going forward, I’m going to focus it pretty much exclusively on advice for getting data scientists and the job search process and so forth. The website is markmeloon.com. And like we discussed, there is already a report on LinkedIn that people can get. There’s also a blog post on building a resume specifically for data scientists, but more and more I’m going to be using that as my central hub for where to follow my writing and so forth.

As far as asking me follow-up questions, my e-mail address is simply [email protected], so feel free to drop me a line, ask me some questions. You can also find me on LinkedIn. Sort of as a way of keeping up with things, I’m getting more and more people pinging me there. So just look for Mark Meloon on LinkedIn, send me a connection request because I accept all connection requests from data scientists, and then you can send me some follow-up questions there and I should be able to get back to you in a day or two.

Kirill: Fantastic. And I have to mention that Mark has been so modest throughout this podcast. He hasn’t even once mentioned this one thing that he’s doing, but I know that he’s working on a course on how to get a job as a data scientist. I’ve had the privilege to preview the course, and a couple of people from SuperDataScience have as well, it’s shaping out fantastically. So if you ever have an opportunity to check that out, definitely look into that and I’m sure Mark’s putting some amazing content into that. Sorry, Mark, if I gave them a spoiler alert, but I just couldn’t not mention that on the show today.

Mark: Yeah, I recently completed that course, just a couple of weeks ago, and what I found is—like you said, I’ve been so busy on Quora and LinkedIn and so forth and I was answering the same questions over and over again so, number one, I wanted a central place to put all my knowledge so people don’t have to scramble all over the place to try to piece together things.

But another thing that I’ve done with the course is what I call ‘operationalized’ my knowledge. There’s 30-something videos in the course, but also along with that, there’s about 30 worksheets as well because I want to make sure to really lead people step by step. It’s one thing for them to passively get knowledge, but then I also want them to take action based on what they’ve learned. That starts off with strategy and this is something where a lot of people, when they’re looking for a job, they just jump right into updating their resume, so I tell them in the course, “No, no. The first thing you have to do is think strategically, come up with how you’re going to position yourself. Are you going to position yourself as an innovative thinker, or someone who can handle the whole end-to-end process?” You want to start thinking about what types of work you want to do, what types of problems you want to solve, and then position yourself as a solution to that.
So asking the hard questions of yourself like, “Why would someone want to hire me? What are my strengths, really?” You know, there’s worksheets where you actually have to do that kind of work. The whole program is actually called breakintodatascience.com and like I said, it’s a sum total of my knowledge of the process that I’ve used to get data science jobs, but then also it’s really meant as more than just a learning experience, but you actually are building up your assets along the way.

Kirill: Yeah, fantastic. Thanks a lot for taking the time to put your knowledge into that, into something of that magnitude. Thanks a lot for coming onto the show. It was very exciting to have you here and sharing some of your wisdom nuggets. I’ve got one final question for you. What is one book that you would recommend to our listeners to help them become better data scientists?

Mark: Yeah, we’ve talked a lot about the non-machine learning aspects of data science, and I’m really a proponent of that. So the book that I recommend that really everyone starts off with is simply called “Data Science for Business” and the subtitle is “What You Need to Know about Data Mining and Data-Analytic Thinking,” it’s by Provost and Fawcett. It focuses on the application of data mining and machine learning to problems in business, which is rarely covered in books with ‘Data Science’ in the title.

The book itself is not organized in terms of algorithms, although those are covered of course, but rather in terms of fundamental concepts. So Chapter III will tell you right upfront what’s the fundamental concept, and then in there they will actually describe some example algorithms of how you would go about solving that business problem.
So I think the subtitle of the book, “What You Need to Know about Data-Analytic Thinking,” is a great description. As I say, most books on data science are about the tools of the trade, and this book is about what you’re supposed to use those tools for. Like I said, I recommend that every data scientist read this book, especially if they’re early on in their process, because it will help them understand whether data science is really something they’re interested in, but it will also provide powerful context for any other courses or books that they read on the topic.

Kirill: Fantastic. I haven’t heard that one recommended before. Could you repeat the title again, please?

Mark: Sure. It’s called “Data Science for Business: What You Need to Know about Data Mining and Data-Analytic Thinking” and the authors are Foster Provost and Tom Fawcett. And it’s an easy-to-read book, too. There are some mathematical details in there, but that’s somewhat minimized just because there’s other books that deal with that well, like “The Elements of Statistical Learning” and so forth. So this is really focusing on the general concepts of how do you apply data science to business.

Kirill: Gotcha. Actually, now that I’ve thought about it, I think maybe one of our guests previously has recommended this one. Yeah, if you haven’t checked out this book and you want to learn about the general process behind data science and data science for business, this could be the book you’re after.

All right, thank you very much. On that note, we’re going to wrap up. Thank you so much, Mark, for coming on the show and sharing all your insights, wisdom and of course the story that you personally had.

Mark: Yeah, thank you for having me on the program. It’s funny that just a couple of days ago, I was having lunch with a junior data scientist and of course I mentioned, “Hey, I’m going to be on a podcast in the next couple of days,” and he asked which one and I said SuperDataScience and he says, “Oh, I listen to that all the time.” So I know you’ve got a devoted following that gets a lot of value out of your podcast, so I’m thrilled to be here. Thanks for inviting me.

Kirill: Amazing. Thank you, Mark, and have a great day.

Mark: You too.

Kirill: All right, there you have it. That was Mark Meloon from Break into Data Science. I hope you enjoyed this podcast, definitely lots of valuable tips and tricks. My personal favourite was some of the advice that Mark gave on setting up your LinkedIn profile. He mentioned just three tips on this podcast: summary section, coming up with the keywords for your title, using a professional headshot, and of course he’s got many more. So don’t forget to head over to markmeloon.com and download the 17-page report on how to set up your LinkedIn profile.

And personally, I think that this podcast in combination with podcast #69, where we’ve got Caroline McColl, a recruiter from Sydney sharing her insights about data science, I think the combination of these two podcasts is very powerful in the sense that it can help you understand better how to pursue the search of a career in data science.

So, with Caroline McColl we talked about overall how to position yourself as a person who’s passionate about data science, as a thought leader in the space of data science, and build a name for yourself, and one of those sub-elements in her list was a LinkedIn profile. And here in this podcast with Mark Meloon, we spoke specifically about the LinkedIn profile plus you can get some additional tips from Mark on how to structure that. So I think together, all of that should be quite a robust framework on how you can get your name out there and start getting those job opportunities that you will be excited to pursue.

So that’s what we had today. I hope you enjoyed it. And as always, you can get all the resources mentioned in this podcast, all the links, and also a link to Mark’s LinkedIn profile at the show notes on www.superdatascience.com/73. Don’t forget to check that out. And of course, if you have somebody who is interested in a career in data science, who is maybe struggling to get those job opportunities, to build their LinkedIn profile, then I highly encourage you to share this podcast with them. It could really help them out. And on that note, thank you so much for being here. I look forward to seeing you next time. Until then, happy analyzing.

Kirill Eremenko
Kirill Eremenko

I’m a Data Scientist and Entrepreneur. I also teach Data Science Online and host the SDS podcast where I interview some of the most inspiring Data Scientists from all around the world. I am passionate about bringing Data Science and Analytics to the world!

What are you waiting for?

EMPOWER YOUR CAREER WITH SUPERDATASCIENCE

CLAIM YOUR TRIAL MEMBERSHIP NOW
as seen on: