Welcome to our newest season of HumAIn podcast in 2021. HumAIn is your first look at the startups and industry titans that are leading and disrupting ML and AI, data science, developer tools and technical education. I am your host, David Yakobovitch, and this is HumAIn. If you like this episode, remember to subscribe and leave a review. Now, on to our show.
Welcome back to the HumAIn podcast. Listeners, today, our guest speakers are Jacqueline Nolis¹, Head of Data Science at Saturn Cloud³ and Emily Robinson², senior data scientist at Warby Parker⁴. Jacqueline and Emily have recently launched a brand new book with Manning Publications known as “Build a Career in Data Science”.
For all our listeners of the show, you may know me as not only a data scientist, but someone who loves education. So this is a phenomenal book, especially in the time of COVID-19 to see how we can make it in the career of data science. Jacqueline and Emily, Thanks for joining us on the show.
Well, let’s start off a little bit to know more about both of yourselves. Prior to this, offline, We were talking about that, Jacqueline, you’ve been to our Galvanize Seattle location, and Emily, you’re in New York. So we definitely have some shared connections, but love to hear more about both of you.
My name is Jacqueline Nolis. I am a data science consultant, so I go around helping companies like T-Mobile, Expedia, with their data science problems. My background is in mathematics. I got an undergrad in math. Masters in math. Then I worked for a few years and I really wanted to help businesses use math to solve problems. And this was before the field was called #datascience. So I went and worked for a few years. It’s analytics and that sort of thing. And I’m like, “I really want to learn more techniques”.
Because everything I know is theoretical math. So I went and got a doctorate in industrial engineering and then I started working as a consultant. And so, for the last eight or ten years or so, I’ve been doing data science consulting for all sorts of companies and leading data science teams. And now writing a book on how to become and grow your career in data science.
And as for me in college, it wasn’t called data science then, but I basically studied very related fields of statistics. And that’s where I started programming in R, went on from there to get my Master’s in Organizational Behavior and then did Metis, which is another data science bootcamp. Did that for three months and went on to Etsy DataCamp. And now Warby Parker as a senior data scientist.
And basically I got interested in data science because, actually, the social sciences, I find the quantitative social sciences is a very good background to lead into data science, because the social science research process is very similar to thinking of a question you want to answer going out and finding that data that can help you answer it, analyzing it.
And then presenting it to anyone, from your professor who has been in this field for decades to someone in another department who maybe doesn’t have any special expertise there. And so learning how to tailor what you share about your results in your methods to the audience. So I thought it was a really good way to set me up for a career in data science.
What I love about, Jacqueline, Emily, what you both share is that you’ve evolved your careers from academia to industry. And you’ve been in the industry even before it was called data science, much like myself. I was doing actuarial science and business intelligence. Data analytics. And now the industry has coalesced into data science. We’ve even seen a lot of courses coming out online, recently on LinkedIn Learning that future-proofing your data science career course launched, which could not be more timely, just like the launch of your book, Build A Career In Data Science. Looking at everything going on in the world today, not only #COVID19, but the acceleration of technology. Why is this timely now to build a career in data science?
So the reason why I would say it’s timely now. So Emily and I, we both go around, we go to a lot of conferences and we give talks, and we actually met each other because we were both giving talks to the same conference.
And when you give a lot of talks at conferences, you get a lot of people coming up to you asking how did you get to where you are? Or, if they’re not data scientists yet, , “I really thought what you do seems cool. I really want to be that”. And we both have 10 points where we’re like, Oh my goodness. So many people like that.
There’s just, clearly, some desire in the world that people are data scientists, or if you’re a junior data scientist, a desire in the world to be one of these senior data scientists, giving talks at conferences and joining the community. And so we just noticed organically that this is happening more than us making some grand observation about the state of the world. I would say, Emily, or feel free to disagree with me?
No. Why I was so interested in writing a book was because it’s really as Jacqueline says a great way to scale up advice, because I still got LinkedIn messages, other folks reaching out, wanting to get started. And I definitely want to help people, but at some point we found it as it’s been better for us to take time to write blog posts, take a lot of time to write this book, really think about this stuff, work with each other. We also interview a lot of people in the books, we get their perspectives.
And that is really the best help I can offer people who want to get into this field. And I do think there’s a lot of people, both Jacqueline was saying and I want to get into it, but then you bring up the current moment also recognizing, okay, how may I become even more valuable to employers? I may end up having to do a job search. What can I do to prepare so that I can be an attractive candidate to different companies?
And one thing that you both mentioned so much is about building that portfolio, whether it’s with a bootcamp program or getting involved in apprenticeship, it’s getting that hands-on tangible experience in data science. We look at it all across the field, whether it’s the portfolio, the simulations, the case studies, the capstones, these are ways to get your hands dirty coding so that you gain that experience and proficiency and scale up the advice as you put it so well, right there, Emily. I’d like to dive deeper into the book, into “Build A Career In Data Science”. Can you tell us a little bit about what are the chapters and what you’re focusing on in the book?
I can take that. So the book was put up into four parts, and the first part is, basically, what is data science? What does it look like at different companies? So someone who’s heard the term, maybe not that familiar with wondering, how do I get these skills? What are the different paths? And chapter four there, as you mentioned, is on building a portfolio. So that’s the first part.
The second part is great. You’ve got some skills, now it’s time for the job search. So how do you find jobs? What does the interview process look like all the way up to negotiating an offer? So that’s the first half. And then the second half of the book, and the third part is around settling into your job.
So what should you expect in the first month? What is our advice for writing a good analysis, putting a #machinelearning model into production. And dealing with stakeholders. And then, finally, the last half is about when you start settling in and this is where someone who’s even been in the field for a year or two might still find a lot of value, because it’s about continuing to grow. So doing that by joining the community handling failure, which is pretty much inevitable when you’re a data scientist going on to a new job.
And then the final chapter is what are the things you can do even after you become a senior data scientist. So thinking about management, independent consulting or being a principal data scientist. Finally, actually we have an interview appendix with over 30 interview questions, example answers, and then also, some notes about what usually people are looking for when they ask these questions and what makes a good answer.
So it’s so interesting because the book is hitting on two core areas of data science: first is just when you’re getting started, and then, when you’re accelerating that growth to grow from, l as you both well put it, especially Jacqueline, from a junior to a senior data scientist, and there’s many things to do there.
You mentioned stakeholders, the community moving up the ladder, leaving the job gracefully. These are just, some of the chapters are covering in the book and I’d love to get to all of them. But before we dive deeper into them and why they’re relevant now, of course, the big topic of the year of the century is COVID-19.
And for us at Galvanize, we do a lot with both consumer and enterprise. So similar to some of the clients you mentioned like Warby Parker, T-Mobile and Expedia, we’re exploring these different companies for data science teams and what it looks like to manage data science during the time of COVID-19. In your experience, what are some of the things that you’re seeing?
So I would say the thing I’ve been seeing a lot just in general is, no one really knows what’s happening. No one, or for the last two months, no one really knows what happened. No one knows what’s going to happen for a while. That we’re just in a really uncertain time. And so, when it comes to employing teams of data scientists, hiring is slow because do you really want to start? you really want to double the size of your data science team.
We don’t know if your company is going to be around in six months, everything’s more uncertain. And so, staffing decisions around data science are being put on pause because, often, data science isn’t the one thing keeping the company running. So things are just a little bit tighter in that regard.
And then, I would also say, similarly, just from a consulting side, a lot of companies just in the same way as they’re not necessarily hiring and growing their staff, should we try and build that innovative next best action model or should we try and do that innovative churn thing, right now in the middle of the virus? Or should we really stick to the basics of what our job is instead of trying to push the envelope? So it seems a lot of hunkering down, is what I’ve been seeing.
And definitely, some within companies, shifting priorities and being prepared to adapt with that. You might have been working on this project that, as Jacqueline says, has been made more ambitious, maybe longer term and not needing to shift to a shorter term need because the direction the business is going is different. So Warby, for example, all of our retail stores are currently closed. So of course, we’ve been focusing on e-commerce and that’s how Warby Parker started. But what Jacqueline was saying and what I’m saying is, it definitely doesn’t just hold for data science.
A lot of companies are putting on hiring freezes in general, except for very critical roles, or lots of teams are shifting what they’re doing. So, for people now it’s just preparing to be adaptable. And also, whether that’s in a job search and maybe looking at jobs that originally don’t necessarily meet all of your criteria, just having to be flexible there or within your work, being like, this team needs a dashboard and it isn’t the most technically interesting work. No, but this can help them right now. It can help make some more money. And it’s okay that I put on pause. It’s a more ambitious model that we’re not going to see a payoff from for a year.
That’s so well put. Whether it’s internally at the companies that we’re all working at today, or for all of our listeners here, I say that, if you’re currently employed in data science, you should be looking for opportunities within, and it may not just be traditional data science. It could be dashboarding, it could be helping other business units. It could be product marketing, sales engineering, developer advocacy, wherever that falls, so that you can help your organization succeed at the same end. Guess what, you get to learn some new skills, and although they’re not all coding in Python and R they’re directly relevant to accelerating your #careergrowth.
So with that in mind, we’ve talked a little bit about COVID, but I’d love to dive deeper into some of these chapters. Especially around enterprises, looking at big companies scale and how there’s so many dynamics with the people and the processes and the systems at play. One chapter that I found really fascinating was about managing stakeholders. And every organization has different stakeholders, but what are some of the common threats that you’ve seen around managing stakeholders?
So, let me just say at the start, when we’re brainstorming the book Build A Career in Data Science, you’re thinking about what chapters you want, and you think a lot of the typical things about data science. Like making analyses, choosing the right language. What about databases? We looked at a lot of that. We have some interesting stuff to say, but nothing particular, nothing worth writing, we are reading a book about our stance on that. Databases or those things like that.
But that’s really important to be a good data scientist, and people don’t talk about it as much as they should, or something like managing stakeholders. This is really something that will make the difference on if a data scientist is successful or not much more than my SQL or PostgreSQL are.
And so, we wrote a chapter for this book about stakeholders, and it’s hard because how you work with stakeholders and really how you interact with anyone who is not a data scientist, that can really be a pivotal part of your job, but there’s so many things that go into that.
So one of the things we do in the chapter is we really break it down into thinking about the different types of stakeholders you have. So, you may have the engineering stakeholders who take the output of your machine learning model that you’ve created and actually deploy it. And you may have the business stakeholders who take the analysis you use a model for and have to make a decision on it. And then you have the executive stakeholders, that have to lead the whole business and look at your team doing something interesting.
It’s , wow, that might be the future. I’d really need to hone in on that. So each one of those stakeholders has a different goal, whether it’s to make their engineering stronger, to make better decisions, to make their company go to a better place in the long term. And how you work with each one of these groups of people really will differ based on who they are and what their goals are. So we break down that a lot.
And then, we also talk about how you think about them as people. We have a great quote from Elizabeth Hunter, who is an executive at T-Mobile, our executive vice president at T-Mobile. We have a lot of discussion around how you think through working with stakeholders.
And it was interesting. As I mentioned before, we have these interviews at the end of every chapter and of course, we had an interview at the end of chapter 12, exactly on working with stakeholders. But it came like communication in general, really.
It came up in a lot of different interviews, and a lot of different side lurbs of a skill that sometimes is lacking a new data scientist isn’t necessarily taught in programs, but is really crucial for having success in the career. And that’s why we wanted to demystify it a little bit, talk about it and explain why we think it’s important. And also as Jacqueline was saying, different strategies that you can adapt.
It’s so interesting because as someone who works on both products and projects, I’ve seen many stakeholders in my career, some of them could be the manager, the software engineer, the data analyst, data engineer, the C-suite, other analysts in the team. They all have different things going on. And you’ve mentioned and shared that depending on what the goals are and who they are, it’s very critical.
One thing I love to do is build personas on who these stakeholders are, or even do simulations. What would the conversation look like? What would a product workflow look like? That could help with communicating effectively. But there’s so much more than thinking about how to communicate. It’s actually communication. What are some of the key communication strategies you’ve seen be effective as well?
So let me preface this, which is, this is really hard. And Emily and I, We did a talk for some students in Duke, and I remember they had this question, how do you get good at working with people? How do you get good at communication? “I don’t know, you just do it a long time”. And the answer I went with was, you just mess up a lot until you remember how you messed up the last time, and then get a little bit better. And you do that for 10 or 20 years. And eventually you’re okay.
That was the best answer I could think of, but it’s really like, how do you do human good? There’s really such a vast question of how you connect with people. So it’s really hard. And besides practicing, doing research, reading books on this stuff and really just thinking about it a lot, I don’t know if I have a really cool one-sentence answer that solves the stakeholder communication problem.
There’s definitely not one neat trick, but I do think Jacqueline you’re selling our chapter a little bit short. We did write 30 pages on stuff like audits, and it wasn’t like learning how to be human, good or failing. Although I do think you’re entirely right.
So one thing I write about is being consistent. Creating a consistent framework for how you share things. I also thought of speaking, adapting it. So I saw a tweet from Oscar Baruffa the other day which said he’s finding himself in the position of a stakeholder some other analysts work and found an eye-opener. And one thing he said was that an email after a couple of weeks with ‘here are the results’ is really baffling. He has all these projects going on, reminding him what they’d agreed on the last meeting, what you did and how to interpret the results.
So that’s really good, basically offering context, but it was interesting because I’ve almost always worked in, Warby is the exception, but at DataCamp and actually I worked in embedded teams. So I was working as an analyst with one team. And actually, I didn’t usually need to do this because they knew exactly what I was talking about because we work together every day. So they knew it was a project, they were working on the engineering side, for example.
So I thought that was an interesting thing to have, you can imagine, just because you maybe master communicating with engineers or when you’re embedded, it doesn’t mean maybe I’ll have to learn new things when you’re working with a couple teams and they are working on stuff that you’re not. So you have to adapt your strategies.
Emily is right. We do give tips and guidance in the book and I recommend you do research. And buy our book, of course. I would also say another thing we talk about in that chapter is just the idea of how you prioritize this work. Because it’s very easy as a data scientist, especially as a junior data scientist, you just do whatever the last person told you to do.
And then, especially as you got more senior and suddenly you have 10 things coming in at once and you have to notice some stuff. But sometimes you really shouldn’t be the person saying, no, it should be your boss saying no, but then your boss is on vacation the way you do. Just having, thinking through a lot of the prioritization and deciding what work to do when that’s really important to good #stakeholdermanagement and something. We covered the buck with them.
That’s excellent. One of our best practices that I work on our data science teams at Galvanize Enterprise is we use JIRA and different Kanban boards to definitely plan out what our product releases look like, what our software stand-ups are on a daily basis.
You’re exactly right. You got to set up those prioritizations to be doing the work, but it’s not all just about doing the work and managing the stakeholders, because eventually you’re going to ship the releases, the products, the models, the software, whatever it looks like. And it’s not always pretty.
One of the biggest phrases we’ve heard from a lot of research institutes is that less than 20% of Fortune 500s have deployed AI or data science in production. There’s a lot of failed products. In your perspective, you have a chapter in the book on deploying #models in production, about how to handle these failed products. What are some of the things you have both seen about why projects fail?
On why projects fail, I would say projects that you might put in production fail. They never make it to production. And projects where you’re supposed to do an analysis and show a presentation to an executive making decision also fail. Failure can come in all shapes and sizes. For me, I find one of the most difficult types of failure is that when you’re a data scientist, you generally have to get people excited about a project before it starts like, guys, we’re going to come in, we’re going to do this great churn model. And it’s going to really change the business, give me funding and people. And you got that. You have funding of people, and then you start working with the data. And it turns out that data doesn’t have a signal in it.
Despite having lots of data, you will not actually have a signal to what the role will be next. Because it’s ramped up. And so sometimes there’s just not actually a single in there and you have to go back and be like, sorry, that money and people didn’t really get us what we wanted. And people hope if they just use a more advanced model on that data, they eventually will find the single in there. But that’s rarely the case. Usually, if you can’t find it with a simple model, you’re never going to find it. And that’s a really big source of failure in the data science field.
And that’s why I do think it’s valuable. Balancing the projects you take on, because sometimes you, like Jacqueline say, you can’t tell ahead of time, if the project’s going to fail, but you can usually tell what’s going to be more riskier than others. And we’re working with data we’ve never worked with before. Are we trying to do a predictive model? are we making a dashboard or something you’re more confident and that’s more analytics work? which often has a lower failure rate.
It’s more about presenting the data that you have. So it’s also worth thinking about, as a team, maybe not taking on only pie in the sky, very high risks, new cutting edge projects and balancing that with things that you’re more confident you can deliver because that can help show people the value of the team. And then, hopefully occasionally, one of those riskier projects does pay off and it will probably pay off in a bigger way. And that can be really advantageous, but not necessarily betting everything on that.
Traditionally, some of the G mafia or failing companies have been known very much about focusing on shipping products. We’ve seen that with both Microsoft and Google, that employees traditionally spend upwards of 80% of their time working on core products that have stable features and releases and updates to dashboards and analytics, but then the other 20% could be these pie in the sky projects. It’d be shiny object syndrome because they’re moonshots.
They might change everything or they might not get there just yet. I’m curious about when a project fails. What’s next? For all of us in life failure is pretty difficult to accept, rejection is pretty difficult to acknowledge. How can we better manage the risk of failure?
I would say this goes back to the stakeholder chapter thing a bit. I was talking about that type of failure before, where you’re like, guys, everyone, please, your attention. We’re going to make a cool data science model with this data. It’s going to be great. Save the company. And it doesn’t. And you have to let them down. That is a lot harder than if the conversation started with, Hey, I think it maybe, might work for us to try and build a #churnmodel, but I’m not sure.
And I’d like to do some research and try it a little bit first and see what we can do. And then if you do that, and then six weeks later, you’re like, ‘we tried it. It didn’t really show any promise. And so we’re going to cut it off’. You manage the expectations. And so when reality sets in, it’s not so bad. So a lot of the work you need to do to handle a failure really starts long before the failure actually occurred. Otherwise, you’re just making it really hard for yourself and trying to let people down.
And here is an end of chapter interview with Michelle Keim. She is the head of data science and machine learning at Pluralsight. She had some really good insights and a couple of the points she had was that, one, making sure to check in along the way and getting feedback as necessary. So it’s not a surprise at the end.
Second, failure is actually good. He wants some failures. I remember when Jacqueline was writing this chapter, I was thinking back and I was like, I’m actually not sure if I failed that much in my career. And I don’t think that’s a good thing. And I wasn’t like, Oh, that’s because I’m the best data scientist ever. Actually I haven’t been taking a lot of risks and I’ve been playing it safe a bit more, doing things I know I could take on, or I had the support person who could help. I got stuck. And so that’s actually been something that I’ve been wanting to do more, pushing yourself a little bit. Because it means you’re learning and you’re growing, and there’s not really a replacement for failure.
And then the final thing too, though, is that companies do have different cultures around failure, and at some places it’s not seen as valuable, you might be punished for it. So I do think it’s worth listening to what Michelle brought up. It’s worth one job searching, trying to understand if that company has a culture of learning and ongoing feedback, because you do want to be at a place where it can be safe and understood that sometimes things do fail. And again, that is good. And especially in data science, it’s pretty much inevitable if you’re taking on certain types of projects.
That makes so much sense because we’re talking about New York City and Seattle and Silicon Valley about the startups where it’s normal to experiment and iterate and go through failure, and failure doesn’t mean the startups go bankrupt.
It just means, “the experiment didn’t work. Let’s run another experiment. Let’s run another test”. But depending on what company you’re working for, these failures do require you to explain the results. So failures still can become wins for the company to explain what was the reason the result happened. Do we need to collect more data? Do we need to reevaluate how the problem is being explored from design thinking? Or are there other hidden variables that may be there that cause this, that we don’t know about just yet.
And as both of you, Emily and Jacqueline shared, some of that starts with conversations with the stakeholders at the start of the project, talking to managers, saying, we want to consider to explore all these things. These are different questions that might help us during our journey, but we’re going to approach it in the scientific process so that we can see will we get success, but if not, it’s an experiment at its finest.
And I would just add on, I would challenge the idea you were saying a little bit about, the idea that startups are more comfortable with failing fast and frequently because startups are lean and exciting.
I’ve worked at small companies. I’ve worked at big companies where we had a massive project fail, but we lost $2 million, who cares. We have a lot of millions. And I’ve worked at startups where things went slightly wrong and the management flipped out because they’re clinging to one chance they might have a break out of the startup world and become a big business. So it really depends on the culture, as Emily was saying. And I don’t know if that, in my experience, it hasn’t correlated as much with size as I would have expected.
Now, thinking about culture. So we are now in a remote-only world that has gone completely digital as a result of COVID. We’re seeing a lot of platforms coming out there. Traditional ones, people are used to working remote Microsoft teams, Slack and Zoom and Google meet, and a lot of other interesting startups out there to work remotely, to mix up what that remote culture looks like. There’s Tandem to work together. There’s Rezo for these online conferences, every startup popping up every other day.
It’s really interesting. But my question I have for both of you is, about remote culture and how to successfully manage or build those workflows remotely. In my experience, I’ve always had a role. The last few years has been a hybrid in-person and remote. So the transition wasn’t too hard.
For me, it’s been overly communicating. It’s helpful because you’re not always seen in a remote culture, but I wanted to hear from your perspectives, how has that transition been for both of you and what are some of the norms that we can do in a remote culture to be successful?
A couple of things. So for our teams, the data science team at Warby Parker, already had two remote members and then four people in person. So it was a little easier for us, we were used to not all being in person, but now of course the whole company is remote. And actually, one tool that’s really helped in data science, which you actually didn’t have before is, I don’t know whether it’s Topal or Tupel, but it’s a screen-share and coding program. You can take over the other person’s screen.
You go into a certain mode or you can be in drawing mode and point things to them, and it’s made it so much better than being like, “go to line 80”. You can type this thing, spell it this way. So that’s been a really nice tool for remotely pairing.
The other thing that’s been fun is we’ve been a tech team at Warby doing things like happy hours or show and tell. So someone did a virtual cocktail hour and just finding ways to connect with people and recognizing, for example, in team meetings that there may be more chit chat at the beginning then there normally is, it’s so building in time for that, because people want to connect and catch up and see how folks are doing.
And then finally, the definite don’t is, “Oh, I was reading this New York Times article about employers that are getting these softwares to monitor their employee’s computers, that will take screenshots every 10 minutes. It makes a productivity measure that’s based on how many mouse clicks and words you’re typing. It’s not productivity at all.
Two, it hugely invades privacy. You’re going to have it installed on your phone and see where you go. So I do think that’s a bad way to do it as a takeover by micromanaging. And if you have a healthy #teamculture you should be more, you should know what outcomes you’re striving for. What success looks like there, trust your team to do the work well, to give them the flexibility.
Maybe they won’t work the normal nine to five hours because they have a kid at home, but trust them to get their work done. And also understand this isn’t normal right now. We’re not just working remotely, we’re working remotely in a pandemic. And having that human understanding that people are going through different stuff. So that was a lot, but the biggest thing is having empathy. Don’t invade people’s privacy and respect that folks probably need more flexibility than they might have had before.
So I agree with everything Emily said, and I would just add to that. So before the coronavirus, years ago, I worked for a long time at a company where everyone was remote. It was a consultant company and it was the best. It was so good. My commute was walking down a hallway. Traffic was the cat and you got worked on. And just as Emily was saying, it was a community where we were on Google Hangouts all the time talking as we needed, but no one was checking your hours.
No one was making sure you did this or that. It was like everyone assumed everyone else was an adult and you just got your work done. And it was great. And we got our work done. We consulted, we made money or whatever. That was great. And currently, it’s a very different environment where everyone’s suddenly forced to do it.
There’s childcare issues. It’s very different, but the idea, as Emily was saying, is everyone gets their stuff done. Fine. That being said, once this virus is over, I could never be the one remote person on a team where no one else is remote. Like Emily was talking about, having two people at Warby doing that. I could never do that because I have real FOMO fear of missing out. And every day would be like, they’re having a party at the office and they’re talking about me and I’m not there. So I would really struggle with that. And trying to solve that problem is a perpetual dynamic thing. But that’s different than perhaps the problem we have right now where literally everyone is forced to be remote.
And when we are remote, it is possible as we’re hearing from all these case studies to be successful, as long as we are tracking and tracing, COVID, not employees. We see all those track and trace apps, but remote is not only for work, as we’re moving into this digital-first or digital-only life for an extended period of time. It’s not only our work, but it’s everything, but the entire work-life balance.
We’ve seen a lot of Instagram celebrities and athletes who are now doing live training on Instagram Live and Zoom and all these sessions. And one of the big missing pieces in the data science life for a lot of engineers has been meetup culture, conference culture. Traditionally this is in person. We’re flying or we’re submitting research or presenting a conference as both of yourselves, as newly minted authors as well.
You go to the conferences and share what communities look like today. In my experience in New York, I’ve seen some of these startups, like the New York Tech meetup, one of my favorite meetups, all about tech, they had in the last few weeks their first Post COVID world meetup. That was all online. It went pretty smooth, and there were about a hundred people there. So it was pretty interesting. But the dynamics are very different. I wanted to hear from your perspectives, what are you seeing in the community today in the data science world? and then, we can even step back on the more macro level.
So I would say first, you’re right. Meetups are important. Conferences are important. What you do in this virus is Twitter. Emily and I we’re on Twitter a lot.
Twitter is all about COVID now. It’s not as funny.
It’s angry. It’s an angry conference that is also very friendly, but angry at the same.
Now, is it about COVID with data science or just COVID in general?
And then how the White House forecasts uses a ridiculous polynomial regression. I remember one of my first jobs out of college was doing forecasting and I remember thinking polynomial regressions were terrible. And now here I am, 10 years later hearing everyone talk about it. And 10 years ago, I would’ve been like, that’s amazing that we’re all talking about regressions together and now it’s not what I wanted.
The model was wrong.
But Emily’s right. It’s hard because everyone’s just talking about their COVID lives and not cool data science stuff, which is a bummer. If you want to talk about cool data science stuff, a great place. We want to talk about how difficult your life is right now, which often people do. So I did my first virtual conference this week. I attended, it was the Women in Data Science, Puget Sound conference, and it was a one day conference. It was really interesting. I thought I had a really cool program to do it, but even still, it’s just hard. It’s not the same as just seeing someone and waving at them in the conference. So it’s been a struggle.
We were planning to do two book launch parties, one in New York and one in Seattle, in April, May. Of course that’s been indefinitely postponed. And we’re still doing, obviously we’re doing this podcast. One cool thing is getting to do Meetups in other cities.
They wouldn’t necessarily pay it for you to come up to Montreal. So that’s one benefit, but I do think it’s harder. I’m looking forward to, maybe meetups innovating by doing things like Zoom breakout rooms, finding ways to help people connect. Because that’s often what I would say to people like a big value of conferences and meetups.
The talks are very valuable, but it’s also the hallway track. Whether it’s catching up with people you know or getting to meet new people, having those interactions, which you can’t really do. It’s 200 people in a Zoom meeting. So figuring out ways to facilitate that, the tech and the meetups, as this goes on, are going to find ways to do that.
And what’s really interesting about the conferences is both of yourselves, I’m following many of the data science conferences and ones that would be in person, that would normally cost a lot of dinero have become free. So you could actually join these conferences online and network.
And it’s always a great time to network, especially since many of the conferences are free today. Register, get the material, learn about the new algorithms, the new techniques. Conferences are such a great way, whether during COVID or life after COVID, to stay relevant on what’s going on in industry. Beyond that, what are some ways that you’d recommend new data scientists or ones that are looking to up their career to get involved in community efforts?
So there’s a couple of different ways. So we mentioned Twitter before. Twitter is a little bit taken over by COVID, which is totally understandable, but there’s still some data science, Twitter, in general. We definitely recommend it to people, especially in the R community, like the R stats hashtag is a very friendly community. It’s a nice way to interact with people, maybe even admired for a while, who are big in the community. So that’s one way. I also am a big component, a proponent of doing public work.
So of course we talked about building a portfolio, to get a job, but even after that, maybe continue doing a blog and that’s a great way or, or speaking, a really great way to get out there a little bit, have people start to recognize your name, know what you do.
Normally in person meetups, I really like speaking, because then people will come up to you afterward and they’ll have something to talk with you about, it’s really nice at a big conference because I’m very introverted. I don’t necessarily need to put myself out there. People may have heard my talk or heard of it and they’ll come up and talk to me about it. Those are some of the ways.
And the final one is we also have advice. If you want to reach out to someone specifically and a couple pieces on that, let’s say you don’t want to get advice from someone. And just a couple pieces of advice there is to check out the work they’ve already done publicly. Maybe they’ve done a blog post that answers your question. And maybe if it doesn’t fully, you could reference when you reached out to them, Hey, so-and-so, I read your blog posts on whiteboard, coding interviews.
I thought it was really interesting. I was wondering if you had 30 minutes to chat about some follow up question on that, because I do think you don’t want to necessarily just reach out with ‘can you be my mentor?’ Or ‘can I pick your brain?’ You want to give people a specific time box, ask and also recognize now some people have more time, but some people have a lot less time.
So this may not be at all a good time for them for someone they don’t know. But on the other hand, maybe they do have some free time and this could be an easier way to reach out to folks who aren’t in the same city as you, because everyone’s doing everything virtually anyway.
I cannot agree more like this. So much to what you just shared there, Emily, myself as well, I publish in Towards Data Science on Medium and have a lot of content out there. And when I’ve had people reach out to me and they say, I read your article and I have some more questions. And this is the question I have. I’m saying, wow, this is so awesome. You read the article and you have questions. I give my time, free time. I love this. This is so awesome.
So I totally say that to anyone when you’re starting out your career or you’re building your career, that someone you’re looking up to, you have a question on read their body of literature or work and be sure to invite that into the conversation when you reach out, it is such a fantastic way to build rapport with your new audience.
And speaking of just overall audiences, we are in such an incredible, unprecedented, I’m going to use that word, time, as it’s the most popular word of the year beyond COVID, where everyone is experiencing potentially burn out. We’re in a world where there’s really no more work-life balance or it doesn’t seem so as much.
And I know for me, one thing I’ve done to minimize burnout has been working out daily. And I brought back the beach body workouts that I used to do in the early 2010s now to a daily practice to keep myself consistent, not only in shape, but to have that balance, what are either some things that both of you are doing today to minimize burnout.
Sure. So, it used to be just as Emily was saying. It used to be what I would do in my free time, it’s more data science stuff. I’d write blog posts. We wrote a book together, make fun projects, make a weird thing about generating neural networks for pet names, I do all of these things as, Oh my free time, I’m going to do more work.
And in the last a couple of months, I just hit a point in my life where this has to change. I need an actual break. This is burning out. And so I’ve picked up art. So I’ve been doing a lot of watercolor and oil pastel, and it’s been nice to just have something that is totally not tech to put a little bit of my heart into.
And well, it’s actually surprisingly challenging because I noticed my brain is trying to do the same thing I do with a blog post and improve my content. Is this art going to generate the most likes on Twitter? At what point can I sell this art? And it’s really forced me to recognize how much I do this and dial it back.
So I really struggled with that a little bit, but that being said, I’ve been really happy with it. And it’s funny because on my Twitter feed used to be all my data science take hotcakes or whatever. Now it’s just filled with my art. It’s a nice change.
And for me, you mentioned exercise. So I came out in mid-March to Utah to stay at my parents’ place. And one thing that’s really helped me is mom, so I’ve been keeping about New York hours. So around 4:35, usually I’ll go for a walk and we have a dog out of here and I’ll go for half an hour, 45 minutes, which has been a really nice way to signal the end of the day, get some exercise, get outside and have fun.
So that’s something that I found really important and definitely helped me. And that can be harder in some places than others. A lot of my friends in New York City don’t feel comfortable doing that because maybe they’re on high floors in their apartment buildings. And so it’s either take 10 flights of stairs or take an elevator and then, the streets are crowded. So I definitely do, and you have to adapt somewhat to where you are, and the circumstances. And so maybe for someone else, doing a yoga class instead, and that virtual yoga class for half an hour. But it’s nice to have something that signals the end of the day.
One of the biggest things that I’m so loved that you just brought up, Emily, is that if you’re in a big city, if you’re in a big building, we don’t do this enough in America, but everyone takes the stairs. It’s such a great opportunity to get exercise. Now I don’t do it often enough, but sometimes I’ll take at least 20 flights of stairs to my building. And it is a great workout, you do a little bit of a core. You do a little bit of cardio. You do that a couple of times a day, or a few times a week, it can be helpful, but enough about moving our bodies and moving our minds to different skills like art.
How about moving on? Our final topic we’ll talk about today is, of course, we’re in these unprecedented times with COVID, but often in careers, people think about, when do I move on as in, can I move to a more senior role or can I leave my company? Or if my company doesn’t even exist anymore and I’ve been furloughed, there’s so many interesting things right now. I want to dive into them a little bit.
So first and foremost, let’s start with the least severe. You’re currently working at a company. This is even prior to COVID and post COVID, and you’re determining what’s next. Should I leave the company? Should I stay? What’s your general take for data scientists?
At the current moment, it’s certainly riskier to leave without another job lined up. For example, you’re not necessarily taking on a huge risk by job searching. It might be harder and you’re getting fewer leads, but if you’re staying at your company, that should be secure.
But in general, a big thing we talk about as a signal is, are you still learning? Because data science is a field that progresses fairly quick and of course you can never know everything, but you want to look back three, six months, a year and be ‘Oh wow. I can do some things I couldn’t do before’. Whether that’s you getting more comfortable with the cloud, you’re a better programmer, or you’re better at dealing with stakeholders. It’s not just technical skills, a lot of different skills, but, are you still learning?
And the second part we talk about, is there anything, and this may be especially important now where it’s harder to leave, is there stuff that you’re unhappy with that you can maybe change? So for example, I remember one of my jobs, I was talking to a boss and I was like, ‘I don’t really enjoy doing this work’. I liked these road analyses of AB tests. I didn’t really think there could be anything done about it, but I mentioned it to him and he’s like, ‘it could be an opportunity for us to hire an intern’.
We actually have the budget. This could be a mentoring opportunity for you. It’s probably more work the first couple of weeks, but then they could take it on. And for you, I’ll have you do some other work. And I would have never thought of that. So I’m glad I raised that with him.
Am I making any assumptions about what’s not changeable? Am I assuming that I can’t do a different type of work? Am I assuming that this project has to go on forever? Am I assuming that the team I worked with is fixed and approaching, and talking to your manager about that and he approaches it not as ‘I hate everything’. You must fix it now, but being honest, if you have that good relationship with your manager tell him, here’s some things I’m struggling with or some thoughts I’m having, is there anything we can do to change that? And it looks like Jacqueline has some more thoughts on this. Jacqueline’s actually been a manager. So she will have some very good insights.
I agree with everything Emily was saying, of course, very quick coffee. No, I think I would just add that there are probably things that can be changed. There often are a lot of things that cannot be changed. And sometimes people aren’t always clear on what can and can’t be changed, and expect you to change things that you can’t change. If Emily’s story where the manager said, the tired intern, you could imagine the manager just being like, well, automate it.
Actually, automating would take eight months and would be more work than this. And then it’s like, ‘think of a way out’. There are situations where people expect you to do things you couldn’t possibly do. For me, I’ve found in my career the times where I’ve had the most. It’s time for me to go when there have been the most of these unrealistic pressures and not in ability to talk about them. So, as Emily said, if you have a good manager, you can talk about these things. If you don’t have a good manager and you can’t talk about these things, that often, for me, has been the indication that maybe it’s not the right long-term job for me.
That’s a big thing that we gave in the book. The manager’s really important. Actually, that’s maybe the biggest signal. If you have a bad manager who you feel isn’t helping you, it’s not psychologically safe to talk about, it’s like Jacqueline was saying, does not understand things, that could be a real signal that it’s time to move on.
And then the final thing is, and there are some things that are just out of your control of the change. Maybe your manager recognized that too. We were talking to someone who works in the military and they can’t install our Python on their computer. So they have to do everything Excel and maybe one day with help from people way higher up that could change. But he is not going to be able to change that because there’s lots of security reasons why they do it. So maybe he or someone else will reach the point being, ‘I no longer want to do a job where I can’t do this. So I’m going to leave the military. I’m going to leave’. Governments are more restrictive places, go to a place where there’s much more freedom and technology because I’ve realized that’s important to me.
I would just say, also, people always talk about having five-year plans and a goal. And, what are your goals? And I find that very frustrating because my five-year plan is probably changed once every six months in my whole life, I never had the rest of it, but there are long-term things you can shoot for in data science.
If you think about the way data science stopped, usually you start as a data scientist, which means you’ve been out of school for maybe less than five years. You’re working. They become senior data scientists. And that means, okay, you are independent on your own. You can be trusted doing analyses, making models, whatever. And then from there, it kind of branches.
And then they could become a manager, then your job is to keep the other data scientists working and make things run smoothly. There is a principal data scientist, which is like, now you are the data scientist who has the most thoughtful ideas about doing the models. And you’re really the technical expert and mentor technical components. Whereas a manager just keeps everyone’s goal aligned, strategy-wise and working with stakeholders.
And then lastly, you could just ditch the system entirely and become a consultant and work as a freelancer, which is what I’ve been doing, which can have a huge payout and huge opportunity, but also is incredibly stressful, very risky, and just almost impossible to do right now, given the virus. So, even when we were writing the book, I’ve made the wording very strong about being careful about this. And now, if it were released, if I were writing it today, I would probably make it even stronger.
Well thinking about strong emotions, but more around excitement in technology today in 2020. One thing that I love is that a lot of research and data science teams have learning Fridays where they cover research or review different state-of-the-art things. And one thing I’ve recently seen is the snorkel dry bell from Google, which is helping with data labeling and synthetic data. And I’ve been so excited about where the world of data is going to go. That’s what excites me today in 2020, what are some things in technology that are exciting? Both of you.
So I want to say, and this is a controversial opinion, hot take. I really do not care for giant tech companies to come out with giant technology and we’re supposed to be excited about it. I find that inaccessible. I often find it misleading. And like they say, technology is really good when you dive in a little bit, it’s actually quite poor. So I do not get excited at all by that. But what I get crazy excited about is, you’re on Twitter and you see someone you follow like, ‘I released this new Python or R package that does this one particular thing’.
‘I would love to do that one particular thing and I will download it and I will try it out’. And so, most of my learning has come from incrementally being excited about people’s own little pet projects and trying to start utilizing them that has come from giant technology companies, giant new technology that’s going to change the game . That doesn’t really change the game.
And so, Jacqueline, what are a couple of, maybe those Python or R packages that you’d be willing to name drop.
I will absolutely name drop the two that have been tickling me the most. I just meant the most fun. We have a whole chapter about this in the book, it’s about pet projects. And pet projects and side projects, we really can help you learn more. And so the two packages that I’ve really just been keeping on toying around are this one’s called rayshader by Tyler Morgan-Wall. It’s an art package where you can take landscapes and make beautiful renderings of them that look photo realistic with filters and fun colors. And it’s just incredibly gorgeous visualization of data or of landscape data. I’ve always wanted to do that kind of stuff and I’ve never done it.
And so, the other package isn’t much funnier, it’s Ryan Timpe. It’s an art programmer who made a package that you can turn anything into Lego bricks. And so you can make rafts out of Lego bricks and Lego portraits and stuff. Both of these things are really funny. Funny and interesting and I really like it. It really just invigorates me to take therapy projects, R packages, and then use them to make new things and make my own art. And it keeps me engaged and thinking about programming and data science in a way that Google releasing part II, it doesn’t connect to me as much.
I really love seeing new projects, new things people are doing. But what I get very excited about, too, is when folks start sharing their side projects or blogs, or sharing some of their work, it’s cool.
Now, Julia Silge is doing it. She just started our studio studios, an engineer, she’s been live streaming every week. Her analysis of the newest Tidy Tuesday dataset, which is a project that releases a new data set each week. And she’s been using it to show off the new tidy model system for packages, for modeling in R. So I don’t know. I really enjoy seeing people having fun, like Jacqueline says, folks posting the ratio thing, someone did that, a ratio thing. And he put our book in this spinning thing with a tree. And that was really cool to see.
And the final thing I’m excited about is also more diverse voices and a wider diversity of people being the ones doing that sharing. And there’s been more groups coming out that support different underrepresented groups. It is more recognition of the importance of having diversity, and also the need to be thoughtful that it won’t necessarily just happen. We’ve seen more and more folks realizing it’s not just gonna naturally happen.
It was maybe one of the Python maintainers who talked about the difference between opening the door and inviting someone to dance, just because you have the opportunity to open. You also sometimes need to put in some effort to extend the invite, to let people know that they’re explicitly welcome and wanted, because a lot of these groups haven’t historically been excluded, have not necessarily been one wanted, or had negative experiences.
So I’ve been excited to see what’s happening in the data science community and the recognition of how important that is. In writing our book, we have a majority of women interviewees, and that was not an intentional thing we did. I can’t think of a fair amount. There’s some people in #Python or #R, but I couldn’t imagine asking someone nowadays who is active in the R community to name some of the top people in the R community are programmers and not naming any women in a list of 10, because there are so many prominent folks out there. So there’s more to be done with other groups including people of color, but I’ve also seen some meetup groups and other efforts for that. So that’s what’s exciting to me.
And I feel like it used to really be the idea of that. If you were going to be a knowledgeable person on AI and data science, then you had to be a man with a PhD working at Stanford. People don’t believe that anymore, or fewer people believe that, fewer people believe that every day. And I’m with Emily. It’s really cool to watch.
And so tying that all in together to Building A Career In Data Science, what call to action do you have for our listeners on HumAIn today?
My call to action is to try to find a way to help people. That’s why we wrote the book. It was certainly not so we could get fabulously wealthy and retire early. So whether that’s helping someone at your company, whether it’s writing blog posts, answering a question on Stack Overflow, because usually people know more than they think, it’s really easy to underestimate the knowledge that you have, that you’ve probably have a lot to offer. So maybe think about, what do I wish I knew six months ago, a year ago? and maybe try writing about that, mentoring someone in a posting on LinkedIn, whatever. But just finding a way to share that knowledge, because that’s really how this community thrives.
And I would say that my call to action is, first, I know a great book about Building a Career in Data Science. I highly recommend you buy it. And that is at either datasitecareer.com or the fun version of the book. What does the exact same is that bestbook.cool Anyway, that’s the first call of action. And then the other thing,to build on Emily’s point a little bit, a lot of what we do in our book is we point out that actually, what a lot of people take for conventional wisdom is ridiculous.
And you don’t have to take it seriously. This idea that the right way to apply to a company is with a resume that needs to have bullet points under every job, that stuff doesn’t matter. There’s some stuff that does matter, but a lot of the stuff you hear doesn’t matter. Don’t take conventional wisdom and assume because someone told you it has to be true, including us.
Challenge conventional wisdom a little bit. What you have been assuming that’s true, maybe isn’t true. Maybe if you’re working as a data analyst, you could make it and go switch to be a data scientist, maybe your data analyst and being a data analyst. That’s great.
And don’t let people think that being a data scientist is somehow better or different. That’s conventional wisdom. You have to become a manager, conventional wisdom, your resume has to be formatted in an exact way, but where are you to get a job? There’s so much of this stuff out there that people just assume as true. And the more you can stop and challenge it, the better off you’ll be.
Jacqueline Nolis and Emily Robinson, authors of “Build A Career In Data Science” by Manning Publications. Thank you for being with us today on #HumAIn.
Thank you so much.
Thank you for listening to this episode of the HumAIn podcast. Did the episode measure up to your thoughts on ML and AI, data science, developer tools and technical education?
Share your thoughts with me at humainpodcast.com/contact. Remember to share this episode with a friend, subscribe and leave a review and listen for more episodes of HumAIn.