Michael_Berk:
Hello, everyone. Welcome back to another episode of Adventures in Machine Learning. I'm your host, Michael Burke. Ben Wilson is still unfortunately out, but he will hopefully be back soon. Today, we are joined by a special guest, Adam Ross Nelson. He started off sort of in a non-technical capacity working in communications, then law, then got his PhD in philosophy, and has finally made the amazing transition to data science and is now a career coach. Adam, would you mind introducing yourself and explaining why you are famous?
Adam Ross_Nelson:
Sure, Adam Ross Nelson here. As you indicated in the introduction, Dr. Adam Ross Nelson. And yeah, I was formerly a lawyer. So if I'm famous, and maybe I don't know if I accept the premise of the question, but if I'm famous, it's probably for changing careers so many times. I've worked as a lawyer, I've worked in education administration, my first job ever, ever, ever. was teaching English as a foreign language back in the very late 90s. And one way or another, I've been in education a lot over the years. And eventually after the PhD, actually it's Doctor of Philosophy in Education Policy. So it's really a policy degree that I did for my PhD program. After that, I teched up, transitioned into data science, and I still do data science work. I do data science work and career coaching. And consistent with my background, a lot of my data science work is teaching others how to do data science. I do a lot of corporate training. I found that niche. I found that niche. I'm very fortunate to have found that niche. I enjoy that work a lot. And I also love, love, love helping others transition into data science. And I focus mostly on working, by the way, with mid and late career professionals, so established professionals in some other field. who are interested in transitioning into data science because that was my story. That's how it worked for me. I was in my late 30s before I transitioned into data science.
Michael_Berk:
Did someone help you or was this also a self initiative?
Adam Ross_Nelson:
Oh gosh, that's a great question. A little bit of both. Now, I also would say I didn't make a concerted effort to transition into data science. I was working towards the tail end of my PhD program and then also after my PhD program. I was working at the University of Wisconsin. I was doing advanced analytics for the Division of Enrollment Management there. My boss and a lot of my coworkers started referring to me as the data scientist. And I thought to myself, where is this coming from? Why are people calling me a data scientist? That's not what I trained for. That's not what I aspired for. I'm not even sure I fully knew what a data scientist was at the time. And then some very, very caring individuals, mentors, close friends, started sending me data science position descriptions and said, you should apply for this. You should apply for that. And eventually there was one that really, really, really matched my sweet spot. And this is a strategy that a lot of folks will use, transitioning into data science, especially if they have a previous career. They'll find a role that calls for both data science advanced statistics and analytics, but also deep domain knowledge. So I had extensive domain knowledge in enrollment management, college university student data, national nonprofit organization that works extensively with student data. was looking for a data scientist, a national nonprofit organization based in DC. I interviewed with them, and they hired me. And I was officially, at that point, I was officially a data scientist. But one of the things I really like to point out is I was a data scientist even before I knew I was a data scientist. And there's so many people out there like that. One of my favorite examples of this is Ada Lovelace. Do you know about Ada Lovelace, Michael? Okay, so Ada Lovelace is often credited as being one of the first computer programmers. Like, a couple hundred years ago, I think it was. Oh, I forget the exact dates, I'll have to look it up. Maybe before we're done here, I'll check my notes and I'll get the dates on that. And read the Wikipedia article on her, and she talks about one of her most important quotations in the Wikipedia article. is she talks about how computers can mimic, potentially mimic, all of the mathematical representations of music and then reproduce music on its own. And so essentially, she's talking about data science work, right? Even before computers really existed. So she, too, was a data scientist even before we even knew what data science was.
Michael_Berk:
Yeah, there's tons of people that come in, like are working a specific job that requires domain expertise and then someone gives them data and they're like, holy shit, what do I do? And then they have to self teach themselves a lot of these concepts. And I have tons of friends that have become very successful through this process of trial by fire and Ben actually is one of the one of the classic examples. I am sort of an example as well, though I sort of came from more of a programming background. But yeah, that's super cool. And one of the points that I wanted to ask you about, so after you've made this transition, it must have been sort of slow and steady and you're looking to expedite this process. One of those things that helps expedite this process is having a good professional portfolio.
Adam Ross_Nelson:
Yeah.
Michael_Berk:
So what is that and why is that?
Adam Ross_Nelson:
A professional portfolio can be a lot of things. So I think maybe 10, 15, even 20 years ago, a lot of folks were doing professional portfolios in the form of a website, often with a vanity URL. For example, adamrossnelson.com or michaelbrook.com. And you would place portfolio entries. So you would have representation of various projects. And you would have a I don't know, maybe a couple dozen projects, maybe half dozen projects. I suppose everybody was a little bit different on this. And this is analogous to some of the work that folks would sometimes do earlier. Before websites, personal websites, were very popular in the 90s and 80s, you could build a professional portfolio in a three-ring binder. And it would show examples of your work, your work product. the professional portfolio has exploded. There's really, what I call it, I call it the best, most effective strategy. I call the distributed portfolio approach. So you'll have portfolio entries at GitHub, portfolio entries on LinkedIn, maybe portfolio entries on Medium, and you might even find other ways. I've seen really creative ways of folks sharing. Data science portfolio entries or computer programming portfolio entries just via Google Drive and of course now with Google Co lab as easy it is to share a jupyter notebook via Google Co lab There's another great way to share portfolio entries and the real power of the internet is that all of these entries can cross reference each other and especially considering the notion that that hiring managers, recruiters, potential coworkers may seek you out on LinkedIn or GitHub. When you have cross references, your professional portfolio can grow really quickly and even with a modicum of effort, you can engineer your professional portfolio into a powerful asset that will advance your career.
Michael_Berk:
Why, so it sounds like we're looking to leverage many different platforms and put sort of a piece of you into each of those platforms. Why is that better than having it under one single roof?
Adam Ross_Nelson:
I'm not necessarily sure it is better than having it under one single roof. You could opt to do that, and a lot of folks will still do that, I think. I just think the way the internet works and the way humans interact with the internet, we are using multiple platforms in any given day. By default, if folks are looking to find information about you in... the course of a job search, they might find you on LinkedIn first, or they might find you on GitHub first, or they might find your webpage first. And it's difficult to predict that and it's also difficult to hold someone's hand, hold someone's hand when they're looking for information about you. So like I said, I guess a moment ago, with a modicum of effort, you can sprinkle representations of yourself, portfolio entries essentially, across the internet and really create a, like I said, a strong asset for your search.
Michael_Berk:
Got it. I would also, I mean, this makes a lot of sense, just your argument of how the internet works, I think is sufficient. But also each of those platforms have very different styles. So if
Adam Ross_Nelson:
Yes.
Michael_Berk:
you are looking to showcase your communication skills, maybe medium for a blog, or your coding skills, GitHub is usually the classic. So not only will people increase their likelihood of discovery, but they'll also be able to showcase their different skill sets if you do have a diverse. or distributed portfolio approach, as you've said.
Adam Ross_Nelson:
Yeah, I could give an example of that actually.
Michael_Berk:
Please.
Adam Ross_Nelson:
So one of my favorite projects, and this project actually led to a job offer that I ended up not accepting. I suppose that's another story for another day, but real proof there that this strategy can be effective. So I have one repo which shows coding skills and coding creativity where I generated some fictional data related to birds. There's data related to the length of the bird, the wingspan of the bird, the weight of the bird, the colors of the bird, and the longitude latitude of where we find the bird. Similar to the penguins dataset, similar to the iris dataset. So I generated this fictional data, and then I also demonstrated classification, k nearest neighbors with that fictional data. Did all of that in GitHub. Then I wrote three articles about... that repo over on Medium. I wrote one article about generating the fictional data, another article about k-nearest neighbors, and then another article about just generating fictional data more broadly because I evaluated a variety of tools when I was creating that data. And then I also created two YouTube videos, one that demonstrates again creating the fictional data and the other one that demonstrates k-nearest neighbors. and all of those six pieces, all of those six portfolio entries, so to speak, cross-reference each other and build on each other, but to your point, the point that you were making a moment ago is they leverage each platform's strengths. So the content or the entry over at GitHub is native to GitHub. The three articles on Medium are native to more of a blogging platform, and then of course the YouTube videos are native to YouTube.
Michael_Berk:
Right, and also if you have, so the way Google search works is the more embedded links you have in other websites, the higher the reputation of that page. So if you're leveraging again all those platforms, it increases SEO traffic.
Adam Ross_Nelson:
Yes,
Michael_Berk:
I mean, for
Adam Ross_Nelson:
I believe
Michael_Berk:
canyers
Adam Ross_Nelson:
that's true.
Michael_Berk:
neighbors, you might be a little lower on the totem pole, but for really niche content, you can be in the top results pretty quick.
Adam Ross_Nelson:
I think so, yes.
Michael_Berk:
Cool. So I had a couple other questions, sort of shifting topics a bit.
Adam Ross_Nelson:
Sure.
Michael_Berk:
You developed sort of a personal brand. And I was wondering how you went about creating that personal brand.
Adam Ross_Nelson:
Oh, thanks for asking. I like this question. I did develop a personal brand. I spent quite a bit of time and thought and energy on that. One of the first things I did, very pragmatic, one of the very first things I did is I picked a color palette. And one of the pieces of advice that I don't give my career coaching clients is to develop a personal brand. I feel developing a personal brand for the job search might be over-engineering the process. I definitely provide for clients advice related to establishing a consistent image, which might be a precursor to a personal brand. But for folks who are interested in developing the personal brand or taking that consistent image to the next level, engineering that a little bit further, Starting out with something really simple like picking a color palette can be a Really good way to get started and then using that color palette in all of your platforms The other thing that happened for me in my personal brand. I have the blue vest I and that happened to by accident. I just had this blue vest blue vest. It's sort of a blue puffy vest. We're in podcast Setting here, so I'll just describe it. It's it's a blue puffy vest got it at the mall quite a while ago, back when malls were still a thing. And I just was wearing it a lot in a lot of my videos and then eventually I realized, oh this is now a part of my brand, this puffy vest. It's getting kind of old and tattered, so if anybody, maybe if there's a way to comment on this, on the episode, knows a place to find nice puffy blue blue vests, let me know. I need a new one in order to keep that going. So maybe you'll have it, maybe if you're interested in building a personal brand, you'll have a serendipitous accident like that as well. Just make it authentic. I think that's probably a really good example of making something authentic. I like this blue vest. I wear it a lot. So it's me.
Michael_Berk:
Got it. So get a plushie. Well, I'll get a red vest just
Adam Ross_Nelson:
Yeah,
Michael_Berk:
to switch
Adam Ross_Nelson:
get
Michael_Berk:
it up
Adam Ross_Nelson:
a red
Michael_Berk:
a bit.
Adam Ross_Nelson:
vest, find some sort of signature clothing piece, but probably the more meaningful piece of advice here is just get started with a color palette and then find, my other piece of advice is find others who are more experienced in giving advice on how to build a personal brand than I am. Because I always try to stay in my lane and I'm happy to share my successes. but I don't want to deviate too far out of my own lane there.
Michael_Berk:
Got it. Well, no doubt a large part of creating a personal brand is developing content. And I saw that you have a bunch of blog posts, and you also have a course. Just in one sentence, I quickly signed up. And it seems like the course offers tips on your resume, your GitHub, and LinkedIn, sort of building that professional profile. Would you mind sharing some wisdom in that course? I'm sure people will still sign up for it, but just
Adam Ross_Nelson:
Yeah,
Michael_Berk:
some
Adam Ross_Nelson:
no,
Michael_Berk:
quick hitters.
Adam Ross_Nelson:
I don't mind giving a little bit of a preview. Absolutely not at all. One of the things I enjoy talking about is how you can make entries for your professional portfolio. And again, entries for a professional portfolio might be a GitHub repo. It might be a post on LinkedIn. It might be an article on Medium. It could even be contributing to Wikipedia. So contributing to Wikipedia might be a component of your distributed professional portfolio. So quick hitters, four pieces of advice on ways to add to or enhance your professional portfolio. And I think these four pieces of advice are really good for programmers, folks who are building a programming portfolio, but certainly very good for data scientists as well. Number one is make a Rosetta stone. So if you're not familiar with Rosetta stone Rosetta stone is Archaeological artifact found with three different languages three different ancient languages and it helped us understand How to read these three languages was very important for? For archaeological research if you're not familiar with Rosetta stone just Google Rosetta stone check out the Wikipedia article And that's a really interesting topic. But you can extend that analogy by creating coding Rosetta stones. So you'll have examples of Python, R. For me, I'm usually doing Python, R, and Stata side by side. And when you have examples of the same code crosswalked, so to speak, you could also think of it as a crosswalk. Side by side, you have a really nice. I think, portfolio entry. How do you create a Rosetta Stone? Well, you probably have an existing project. If you're a data scientist or an aspiring data scientist, you probably have existing projects in R or Python. Well, replicate that exact same project in the other language, then write about it. Super simple and easy way to create an entry for your your data science professional portfolio, create a Rosetta Stone, and reuse existing work that you've already done. Number two is make a cheat sheet, or sometimes it's called a quick reference. I actually don't love calling them cheat sheets. I prefer calling them quick references. And if you're not familiar with what a cheat sheet or a quick reference is, I think we'll probably try and get links to the articles and the show notes. But I give examples to a couple different cheat sheets or quick references. One of them is a quick reference. It's a one pager, 8 and 1 half by 11. It's landscape. And it shows, let's see here, about nine different, eight or nine different functional forms with regression modeling. And then it's a quick reference on how to interpret the coefficients on each of those functional forms. So that's number two, build a cheat sheet or a quick reference. And if you're not familiar with Cheat Sheet or Quick References, Google something like Python Pandas Cheat Sheet or R Cheat Sheet, and you'll get examples. Build your own. Number three, write an article about software that you do not like. Now, this one's a little bit risky. You have to do this one with grace and poise. Remember that there are real people on the other end of whatever tool you're writing about, and you might not be liking it. Probably one of the best examples of this is in Stata. I transitioned from Stata to Python years ago. I still use both. But in Stata, it's very, very, very easy to move columns around in your data set. So if you have variables, you can easily move this set of columns to the far left or the far right, whatever. In pandas, that's actually kind of difficult. It's kind of clunky to do that in pandas. So I wrote an article about why I don't like that, why I think that should be a feature request. And I also offered some solutions that might go a long way towards simplifying that in your work. So write an article about software that you don't like, but do that with grace and poise and empathy and compassion for the software developers who built the project originally. And then number four. And actually I have lots of examples on this. Maybe we can talk about those next. Number four is to contribute to another project. So contributing to an open source project is one of the best ways I know to create a data science portfolio entry, build or expand your portfolio. And actually I'll drill down on this one just a little bit more. There are at least five ways to contribute to an open source project. One, you could find a known bug or issue, fix it, do a pull request. Two, you can revise or update the documentation. So you wouldn't necessarily even need to write new code. A lot of open source projects are looking for contributors on the documentation. Three, you could write a new tutorial and submit that to the tools documentation. A lot of tools are looking for that form of contribution as well. propose and then develop a new feature. And then actually I have five of these, five, I misspoke, five. And then the fifth one is troubleshoot existing bugs or issues. And by troubleshooting existing bugs or issues, see if you can find an existing bug or issue and then replicate it. And then write notes, contribute to that issue with your own documentation about how you were or weren't able to replicate the bug or replicate the issue. And that can just help others. as they work through solving the problem or devising a solution.
Michael_Berk:
Got it. So those are a lot of easy and sort of quick entry points into getting into these open source communities.
Adam Ross_Nelson:
Yes.
Michael_Berk:
How do you think about balancing technical load versus street cred? So a lot of these, for instance, like changing documentation, people don't know if you can code, but that is definitely useful for the practitioners that use that code. So what are your thoughts on the value of updating documentation versus trying to develop a full new feature? Obviously the full new feature is a lot harder, but theoretically that would give you more street cred, right? So, I'm going to go ahead and show you a little bit of the full new feature.
Adam Ross_Nelson:
Yes, I think so. I would say for, especially for folks who are earlier in their career or earlier in their coding journey, make sure that one of my big pieces of advice in general in life even is just be authentic to yourself. So if you are indeed, if you do an objective, honest assessment of your skills, abilities, and expertise, and if you're not ready to build out an entire new feature, don't do that. I think that could be a mistake. You're right. That might be worth more credibility. However, we're not doing this purely for the purposes of building our own credibility. We're also looking to make a positive contribution to the community. And that's one of the things employers are looking for, are folks who are willing to make a positive contribution to a community in, by many definitions, an employment setting, a corporate organization is a community of sorts. And if you can demonstrate that you can make meaningful contributions, whether very, very sophisticated coding or really, really, really helpful documentation. If you can demonstrate that you are willing and able to make those contributions in a community setting, that is going to be appealing for employers.
Michael_Berk:
Yeah, and I know a lot of people look at how people submit PRs, how they respond to feedback. That's very telling of how they'll be in the work environment. So if you do go in with sort of an open and kind mindset, that can be very helpful.
Adam Ross_Nelson:
Yes.
Michael_Berk:
I was wondering, though, if you have specific libraries or projects that you suggest contributing to.
Adam Ross_Nelson:
I do actually have a list here of projects that are accepting open or accepting community accepting contributions from the community and in my list I have a couple lists and we might be able to get links to those in the show notes again as well. In putting my list together I prioritize lists or I prioritized packages that have a contributor's guide or have an obvious point of entry, so to speak, into working with that particular tool. The first one is Scikit-learn. Believe it or not, Scikit-learn, I believe, is a really great place for folks who are looking to start contributing to open source projects. They have a contributor's guide. And in their GitHub issue tracker... They have three tags which you can look for if you're just getting started. One is called good first issue. Another one is called easy and another one is called help wanted. They also have a fourth take called documentation. So in the issue tracker and the GitHub issue tracker, you can search for those particular tags and contribute to scikit learn. What, what, what. What a fulfilling experience to contribute to Scikit-learn as a data scientist, right? It's one of the main libraries that we use. And having the opportunity to contribute there I think should be personally fulfilling. In addition to, this gets back to what we were just talking about a moment ago, in addition to building street credibility, but really, you're just contributing to the profession, you're contributing to the community. Another one is called Pandas Profiling. Pandas Profiling, if you're not familiar with Pandas Profiling by the way, check out Pandas Profiling as a tool that can dramatically improve or expedite your exploratory data analysis work. Pandas Profiling, I actually, this week I did a webinar for Statistical Seminars DC. And it was just a simple webinar on Pandas Hacks. and I was showing pandas profiling as one of the hacks and technically I wouldn't call pandas profiling a hack, it's a tool, it's a tool all in and of itself, it's a suite of tools really. And the audience, I thought everybody would, I thought many people would be familiar with pandas profiling, but the comments in the chat, we were using Zoom as our platform, the comments in the chat and the comments on microphone blew me away because There were a high number of folks, high proportion of folks who had not yet seen Pandas profiling, and they were just blown away. I remember the first time I saw Pandas profiling, I was blown away. I promise you, it will improve your exploratory data analysis work as a package. So tip on Pandas profiling, but in terms of contributing to Pandas profiling, they also have a contributor's guide. They also use GitHub for their issue tracking. And they have a really well-organized system of tags so you can find issues that you may want to work on. But they have one issue, which I just checked earlier this week, and it's closed for some reason. So I made a suggestion that they reopen the issue. They are looking for a new name. Pandas profiling is sort of generic for a package name. There's some really interesting creative package names out there in the ecosystem. And Pandas profiling, again, is generic. And they were, at least at one point, looking for a new name. So if you're looking for, even if you are just getting started, you could implement Pandas profiling in two or three lines of code, four or five lines of code, using the documentation. Learn what it is. And then hop on over to that naming issue and suggest new names for the package. Great way to contribute.
Michael_Berk:
And just really quick, pandas profiling is sort of an alternative to bf.describe. It provides a full profile report of the data, which is really helpful.
Adam Ross_Nelson:
Yeah, you could think about pandas profiling as if df.describe, df.info, and the seaborne pair plot had a family and made babies, they would have, pandas profiling would be probably the 10th or 12th generation from those. It's very advanced. And you're right, df.describe would be the, pandas profiling can replace df.describe. I think.
Michael_Berk:
Got it. So those seem like some great packages. Also, if you want to help one of the worst documented projects out there, stats models always can use some more examples. But yeah, that's really cool. Have you ever personally contributed to any of these projects?
Adam Ross_Nelson:
Yeah, I contributed to a couple projects. I've actually lost track of exactly the contributions I've made. But if you search for me in the issue tracker, you'll find me. Actually, no, it's coming back to me now. One of the issues was pandas was having difficulty with saving and reading Stata data format. And I contributed to the discussion. I didn't contribute code to that, but I definitely contributed to the discussion, was able to help demonstrate the errors, reproduce the errors, help other developers solve that issue. And give me a moment, maybe I'll remember the other one that I did too.
Michael_Berk:
Sounds good. Well, in the meantime, I wanted to chat a little bit more on the career coaching side.
Adam Ross_Nelson:
Sure.
Michael_Berk:
Often jobs have a minimum requirements list. And it seems to me even after interviewing people and like defining those minimum requirements, that they're pretty arbitrary. It's like X years of experience, comfortable in this expert in this, knows this and knows this tech stack. accurate or minimum job requirements in your experience.
Adam Ross_Nelson:
I think it can vary from organization to organization, and even within the same organization, it can vary from job description to job description. Sometimes they're really well written, and sometimes they're maybe not so well written in haste. You're probably familiar with, and many of the listeners are probably familiar with, we're looking for something like this. We're looking for someone with 40 years of experience in AWS. Well... Good luck with that because AWS hasn't been around for 40 years. So that's the classic example. True story. I have a true story. So my first data science job is I told that story about how I transitioned earlier in the episode, how I transitioned into my very first data science job. To elaborate on that, I literally went to the interview and I said, I don't meet all of these requirements. And I also said, I also said, I think that as written, this position description isn't well matched, does not match my experiences and background very well. But at the same time, they had flown me out for the interview. But I wanted to be authentic. I wanted to be myself. And so I said that to the hiring manager. And the hiring manager said to me, you know what? We anticipated that. The hiring manager explained to me, I think this is an experience that others may have had, similar experience. The hiring manager said, you know what, we anticipated that. We overstated the job description intentionally. And we also have, this was the real clincher for me, was we also have plenty of people in our organization that know and have. extensive experience with the technologies and the tech stack that we described in the position description, we can teach you and we can train you on that. We're looking for the data science expertise. I was the very first data science hire that organization made. They said we don't have the data science expertise, so if we hire you, we're hiring you for the data science expertise. The rest of it, it's a wish list. So to answer your question directly, I think it varies. I think some position descriptions are really well written. Some position descriptions are not so well written. And well, one of the advantages of having a career coach, by the way, or a career service agency, is the career coach or the career service agency can contact the employer who's listing a job description and ask those questions, ask those really difficult questions. And we can say, We can contact the employer. Sometimes we get through, sometimes we don't. But we can contact the employer and we can say, hey, we're working with one or more clients who might be interested in the position that you've listed. We've noticed this and this and this about the position. We've noticed you're asking for more experience than is actually possible with this particular tech stack. Could you help us understand what you were thinking when you were putting this together? And then we can pass that information back on to our clients. And That is one of the, and then we can do that on behalf of the client so the client doesn't have to, A, doesn't have to do the work, and B, also doesn't have to reveal who they are too soon in the job description, and in the job search process for that particular job description. Those are some of my favorite conversations, speaking with employers and hiring managers about their job descriptions. What were you thinking when you said this?
Michael_Berk:
Interesting, yeah. And how often is subject matter knowledge super important versus just a technical base, would you say?
Adam Ross_Nelson:
You know, subject matter, no, I'm going to attempt a generalization here, and this may come back to, I may disappoint myself in this, but my attempt at generalization is that for smaller organizations and in organizations who are just starting out in expanding or initiating their data science capacity, the domain knowledge is crucial, right? Because they really need folks who can interface with their existing... personnel who eat, sleep, and breathe that domain knowledge. Perhaps at larger organizations, Amazon, Facebook, Netflix, YouTube, Google, the technical knowledge might be more valued because they're going to be in a position to teach you the domain knowledge that you need in order to succeed. And they may not even need to teach you the domain knowledge. They may have entire teams who are dedicated to providing the interface between the technical personnel and the operational or the business personnel. So there's a general observation that may or may not hold true. I don't have empirical data on that, but that is also consistent with my personal experience and consistent with the experience, the anecdotal experience that I know many others have encountered as well.
Michael_Berk:
Yeah, that's consistent with mine as well. We had a guest a few months back who had a pretty impressive academic resume. And as he went through that PhD process, it sort of removed a bunch of jobs that he was looking at prior because he was so specialized and so technical that only a certain group of organizations could leverage that skill set. And they were all like the giant Samsung research, blah, blah, blah. So yeah, that. That makes a lot of sense.
Adam Ross_Nelson:
That was a really interesting question. I'd be interested if folks reach out if you have thoughts on that particular question too. I'd be interested in what others have to think on that.
Michael_Berk:
Yeah, please. OK, cool. So we sort of talked about what you should do prior to any interviewing steps. But let's say you somehow meet the job requirements, amazingly, and you have sort of caught the eye and have gotten into the interview process. Do you have tips for looking like you know what you're doing?
Adam Ross_Nelson:
That's a tough question, I think, because one of the things I want to avoid advising anybody ever to do is trying to fake, fake yourself. The advice out there is fake it till you make it, and depending on the context that advice can be sensible. But make sure you're open, honest, and direct, and be your authentic self in the interview process. Probably a really nice piece of advice to give in response to that particular question is, and this is going to be hard to hear, especially maybe for folks who have only had a few interviews, or maybe in this scenario you just described someone who's at an interview, they really want to look good, and it's the only interview they've had in a long time, or maybe it's their very first interview. My piece of advice for you is, even though you haven't had an interview in a very long time, or even though this is your first interview, there will be other positions. If you're not the best fit for this position, that will become, hopefully that will become apparent to you and or the employer through the search and selection process. And if you don't land this position, there will be other positions in the future. if you keep working on the search. There's a similar scenario. Recently just onboarded a new client, and this particular client is in a position where they have the luxury to be very picky about the positions that they're going to apply for. And the question was, what happens if I see my dream job tomorrow, and we just got started, right? We're getting, right, we're doing onboarding, right? What happens if I see my dream job tomorrow? Should we hustle and apply for it? And my answer was, well, it depends. And part of my answer was very similar to the advice I just gave. Part of my answer was, even though you might think the position you're seeing tomorrow, the hypothetical position that we're seeing tomorrow, that might be approaching dream job status, Even though it looks like that might be a scarce resource, I promise you there will be other similar positions in the future. The quick search on any job board for data scientists will support me on that. So don't try to be something you're not for the sake of any one given position because the number of positions out there is really high.
Michael_Berk:
Yeah, I think I'll echo that. I tend to be a pretty positive person. And so being authentic for just authenticity's sake is a compelling argument. But on the more cynical side, you also tend to perform better if you are authentic and play to your strengths. So if you are a quick thinker and can rely on iterating throughout a problem, do that. If you have a ton of knowledge, rely on spitting out some Rolodex of how gradient boosting works or whatever you're doing. So definitely playing to your strengths is one point. And then another point is, it's really important to remember that the interview process is only three to five, whatever hours long, and they can't test you on everything. So
Adam Ross_Nelson:
Right?
Michael_Berk:
you have a lot of ability to steer where the question goes. And
Adam Ross_Nelson:
Yeah.
Michael_Berk:
classic machine learning questions are, we have this data set, It has, let's say, a categorical target. How would you forecast it with this requirements of this frequency, blah, blah, blah? If you just know five categorical models relatively high level, but then have one that you can go really deep into, that's all they care about. They won't care about cross-referencing super deeply between x versus y versus z. If you can just go super deep into the one thing that you chose. they often don't know that you don't know the other things.
Adam Ross_Nelson:
Yeah, yes. I actually have an example for you. I love your categorical model example. I remember once in a technical interview the problem related to classifying categorical, and I think there were like five categories, and they all had to do with transportation. Three of the five were... I think land transportation and two of the five were not land. They were some other form of transportation, air or boat or something like that. So I proposed, and I just really didn't want to do, I can't remember why now, but there was another nuance to the problem that really just made me not want, this was probably by design on their part, made me not want to deal with the multi-class classification. So I said, hey, let's look at the qualitative nature of these. These three are land-based. These three are not land-based. Let's recode everything to a binary. And then let's do binary classification. And I think I won. I think I won the interview. I got that offer, too. So that went well. And I think it's a really good example of what you just described. Another piece of advice, how do you steer those conversations, as you just suggested? And then also going back to your original question, I think was along the lines of, how do you make yourself look like you know what you're talking about? Well, probably the first way you make yourself look like you know what you're talking about is know what you're talking about. Make sure that you know what you're talking about. But one of the biggest pieces of advice that I give folks when it's time to get ready for an interview or when you're in an interview is ask questions. One of the things I notice when I do... mock interviews is candidates will be reluctant to ask questions about the questions. Folks on the podcast can't see me doing my air quotes here. But you need to ask questions about the questions, which I think is an effective strategy to achieve the objective you just described, steering the conversation in the direction that will let you address your strengths. But I think the reason folks will sometimes be reluctant to ask questions is they're afraid to look like they don't know what they're talking about. In truth, asking a smart question can be a really good way to reveal your critical thinking skills and to reveal that you do know what you're talking about because you're thinking about aspects of the question which are beyond the face of the question. And that's a really important skill that I think you need to demonstrate in the course of your interviews. So, one of the best ways to make yourself look good in an interview, counter-intuitively, is to ask intelligent, meaningful questions.
Michael_Berk:
Yeah, I could not agree more. The most recent interview set that I did was surrounding experimentation. And very few candidates have the technical depth where the interviewer just has no idea what's going on. And often the interviewer, even if they don't know as much as the candidate, they can sort of sniff out that they're BSing.
Adam Ross_Nelson:
Okay.
Michael_Berk:
And what I really value is seeing where the candidate is at and seeing how they go from there. So you. reach the limit of their technical knowledge and then see how they use information in the question to get creative because often we're working on frontiers or at least applying methods to new applications. There's not just a plug and play solution. So if they can go beyond their technical skillset, hopefully they can go beyond whatever skillset is known by the general public
Adam Ross_Nelson:
Yes,
Michael_Berk:
as well.
Adam Ross_Nelson:
yes, I can build on that. Another sort of piece of this, I'll give listeners to the to the episode permission to say this in an interview. Say something like if you reach the limits of your technical skills or even just your your encyclopedic dictionary knowledge, so the trivial knowledge, right? Say, this is an aspect of the problem, I would have to research further in a practice in a practice setting on a professional setting. What you can say in the interview is, if before I could offer a full complete solution here, I would have to research that aspect further. Oh, and then if you know something, if you know of a resource that will provide the answers you're probably seeking, name that resource. And then that will also show that you're familiar with the range of resources that are available in the field.
Michael_Berk:
Yeah, that's a really good point. And it's important to remember that when you're working the job, you have internet and you have time. So
Adam Ross_Nelson:
Right.
Michael_Berk:
no one expects you to be building solutions in a 45 minute block with someone asking you questions. They want to know your process and how you would approach the problem usually.
Adam Ross_Nelson:
Yeah, I think that's correct. And I know that's correct.
Michael_Berk:
Cool. So one final topic before we close out. You made the transition to data science sort of late in your career. What was that process, and do you have tips for others looking to do the same?
Adam Ross_Nelson:
So earlier in the episode I spoke a little bit about how I was a data scientist before I even knew I was a data scientist other folks were regarding me as a data scientist and The piece of advice that story implies I think is if you want to be a data scientist and you're already doing the work of data science Start calling yourself a data scientist The sometimes folks will think that that is dishonest or disgenuine when you call yourself a data scientist merely because your employer doesn't call you a data scientist. So some HR database somewhere lists you as a senior analyst. And just because that HR database is not listing you as a data scientist. doesn't mean you can't call yourself a data scientist. If you're doing data science work, the best way to be a data scientist is to do data science work. So that's advice that I probably wish someone had given to me earlier on. I think maybe I would have become a data scientist sooner. Who knows? Hindsight is always 20 and 20, and I'm not upset about how my career unfolded. I'm very glad that I was able to become a data scientist when I did. But I'm also very thankful for the work that I was doing before I became, technically, according to the HR database, a data scientist. Hopefully that answers the question. I'm not sure I fully answered the question there. More of a walk down memory lane, a chance to reminisce about the transition.
Michael_Berk:
Got it. So that was your experience. And then do you have any pieces of advice other than just changing your title potentially a bit early? Other pieces of advice for getting from your current state, let's say sort of later in your career, over to the data science career path.
Adam Ross_Nelson:
And I don't say changing your title potentially a bit early. I say change your title when it's appropriate, like when there's an objective. And part of that process could be checking with colleagues, checking with friends, checking with supervisors, and say, hey, do you regard me as a data scientist? And you might be surprised at the answer. Other advice the other advice I would give is so what actually here's some advice And this is advice that I invented for myself and or followed from others Who knew more than I did at the time when I transitioned from my? My data science job that really demanded that domain knowledge on college university student data Then I transitioned into big consulting where I worked for large consulting organization, and I was a generalist data scientist at that point. The way I made that transition was I started building my professional portfolio. I started writing about my, I started expanding my professional portfolio beyond GitHub. I started writing about my projects. I started making videos about my projects. I started giving presentations at community organizations about my projects. and then I used that information in order to facilitate my next job search. Here's a real, actually I just remembered this story. This is a true story and I think it might not be representative of many job searches, but I think it's a story about how using the distributed professional portfolio entries as a job search strategy can be effective. So one time I was interviewing for a position, a large organization. They were in a rush. They did not have time for their full, they were looking to fill an unexpected vacancy. They did not have time for their full interview sequence, including their full technical interview sequence. So for the technical, actually there were two technical interviews. For one of the two technical interviews, if I'm remembering this right, There were two interviews, and for the second of the two, the recruiter who worked internally for the organization said, bring a piece of your work, bring an example of your work to the technical interview, and then be prepared to walk the interviewer through the project. And I said, oh, well, I have this pile of articles over here and a couple GitHub repos. and it made the ability to share about that project so much easier. It just, I had to do very little preparation except to review my notes, review the article I'd previously written, and review the YouTube video that I'd previously recorded and published. I just took that to the interview and I said, hey, let's look at this article I wrote. And then, really nice, after the interview, I was able, I had these publicly available links. after the interviewer is able to send to both the recruiter and that interviewer links to the articles, to the videos, to the GitHub repo and say, I was a great chance to chat with you. I'm so thankful for your time and attention in this process. By the way, here's links to everything we talked about today.
Michael_Berk:
Got it. Yeah. Having distributed portfolios is the key.
Adam Ross_Nelson:
That's the way to go.
Michael_Berk:
Yeah. Cool. Well, this has been really fun. In closing, if the audience wants to reach out, where can they find you?
Adam Ross_Nelson:
I am thrilled to connect with people on LinkedIn. You can also find me at adamrossnelson.com. If you go to adamrossnelson.com, right as you land, there will be an opportunity to link out to all of my distributed portfolio, actually. And then also I offer a free career course, and the free career course is designed to really help you with your resume, really help you with your LinkedIn profile, and really help you with your... with your distributed portfolio, and we focus mostly on building GitHub as the backbone or the backend of your distributed portfolio. So if anybody's interested in that, I would encourage you to either reach out to me and ask me questions about it, or find the information at, once again, adamrossnelson.com. And if anybody's looking for some help, and you're interested in a career coach, find me at adamrossnelson.com and we can have a conversation about that too.
Michael_Berk:
Yeah, please do. And I actually signed up for one of the career courses, and there's lots of great material in there. So definitely check out what Adam Ross Nelson has to offer.
Adam Ross_Nelson:
Could I say that I have 2.0 coming out very soon? So
Michael_Berk:
Amazing.
Adam Ross_Nelson:
I would really encourage folks to go ahead and see what we have there now, but 2.0 is going to be even better.
Michael_Berk:
amazing. Well that concludes our episode. It has been Adam Ross Nelson and Michael Burke and thank you everyone for tuning in.
Adam Ross_Nelson:
Thank you, Michael. Bye, everyone.