#25 Data Science for Everyone

Hi, I'm Maggie, VP of Curriculum at datacamp, and I know firsthand how overwhelming it can be to keep up with the rapid changes in data and AI. But staying ahead doesn't have to be hard. That's why our curriculum team has curated the best learning content to help you and your team stay ahead, covering everything from data and AI literacy for busy executives to cutting edge generative AI for.

Developers, and everything in between.

Whether you're upskilling yourself or your team, we make learning practical, engaging and flexible. It just works. Power your data and AI transformation journey.

On datacamp.com in this episode of Data Framed, a Data Camp podcast, I'll be speaking with Jared Lander about building communities in data science. Jared is the Chief Data Scientist of Lander Analytics, a data science consultancy based in New York City. Jared is also the organizer of the New York Open Statistical Programming Meetup and the New York R Conference, as well as being an Adjunct professor of Statistics at Columbia University. Jared has emerged as a key player in the New York City data science community over the last decade and a central mission of all his work is to create safe and welcoming spaces for budding and practicing data scientists of all ilk. How does Jared think about creating this space and put it into practice? How does he make people feel comfortable and at home in a field in which so many intelligent and curious people feel like imposters? What practical and specific considerations are there in creating this home for underrepresented groups? How does he stay ahead of the curve in terms of modern, up to date content and speakers for his meetup and conferences? I'm dizzy with questions. Stick around for the answers. Jared also likes pizza, but as Jared says, who doesn't? I'm Hugo Bown Anderson, a Data Scientist at DataCamp and this is Data Framed. Welcome to Data Framed, a weekly datacamp podcast exploring what data science looks like on the ground for working data scientists and what problems it can solve. I'm your host, Hugo Bown Anderson. You can follow me on Twitter hugobown and datacamp@datacamp. You can find all our episodes and show notes@datacamp.com community podcast. Hi Jared and welcome to Data Framed.

Thanks for having me. I'm excited to be here.

It's such a pleasure to have you on the show and I'm really excited to be here to be talking about building communities in data science. Before we get before we get into that though, I'd like to know how you got into data science originally, so.

I'm going to give you a slightly meandering long winded answer for that, please. I started off in college as a business and communications major and I just took math classes for fun. I liked math, I was good at it and I kept doing very well on the test. It kept getting towards the top or the highest mark on all the tests. So finally, after a few years of my mother prodding me and the math professor saying you can't be getting the highest scores and not be a major, I became a math major. And it just spoke to me. It's like math was fun. It was a great time for me.

And what type of math were you interested in or doing at the time?

I went to a small liberal arts college and it was general math. So of course it went through the calc track, the linear algebra track. There was an OR class I ended up taking without purpose, three stats classes and a stats independent study. Without realizing that stats was going to be a big part of my life, I just took it because it was the next thing you took. In the math department, we took an analysis class and a few different abstract mathematics classes. So it's just a general overview of a bunch of different math.

Great. And what happened next.

So then I did what anybody of a math major would do. I started managing bands. I did that for a few years, working in small bands on my own and working for more established managers and agents on some larger bands. And after a little while of this, I realized that it just wasn't what I wanted to do in my life. It didn't speak to me the way the sciences and the mathematics spoke to me. So I decided to close up shop, get out of the music industry, and I wanted to find any job out there. And I happened to come across a job listing somewhere to build dashboards for fashion companies. And I figured it's a halfway house, the fashion companies is a little like music and the dashboards is technology. So it was a good happy medium for me to transition away from music into tech.

Absolutely. And you'd had experience in some math, some stats. So data, did you have experience in any types of programming or what you'd need to build dashboards?

I had, like many other people, started hacking at HTML when I was in middle school. And then in high school I took a C class and a Java class. Then in college I took one C class and that was my entire experience of programming up to that point.

And then what happened in your life to kind of lead you down this trajectory where you're one of the mainstays in the New York City data science Community.

So after about a year at this company, I was thinking to myself, well, I really like math and I discovered data through this job building dashboards. And I said, well, if you take math plus data, it equals statistics. So I applied to one grad school. I was very much thinking, well, if I get in, I'll go, if I don't, I'll find something else. And I don't want to leave New York City. So I applied to Columbia's master's program and thankfully they let me in. I started off as a part time student, so it took me about two and a half years to go through the whole master's program. And I happened to graduate just before data science was taking off as a term. I happen to get lucky with a few things, with the timing. And I just so happened to write my thesis about New York City pizza.

So what can data science tell us about pizza in New York City?

So I did research on what made one pizzeria more popular than another. And I looked at some data from menu pages, which isn't as big today as it was then, but is a big site for storing the menus of different restaurants and for leaving reviews. And I computed the number of reviews as a proxy and I used that as a response variable in a Poisson regression to figure out what variables that I had access to impacted the popularity. And the biggest factor, it turned out, was having a coal burning oven.

Right, that makes sense. Were there any other features or characteristics that played an important role as well?

Well, the second most important was having a wood burning oven as opposed to gas. And then after that price had a negligible effect. And then depending how you cut the data, being in Midtown was a bad thing for a popular pizza place versus being in the other parts of the city were a little better.

Well, being in Midtown can be, have negative impact on a lot of things, to be honest.

Absolutely.

So this is interesting though, because I think not only does this demonstrate your love of pizza, your love of New York City, but it also demonstrates a variety of different, you know, statistical and data science techniques that go into any project, whether it be web scraping Reviews or using APIs, using statistical inference to extract important features, that type of stuff. So it runs kind of a nice gamut of data scientific techniques.

Absolutely it does.

So after your, your master's, what did you do then kind of leading up to what you, what you do now?

I like to describe it as, I continued to be in the right place at the right time and had some good breaks. A friend of mine who I went to college with and knew I had already started getting into R in grad school. That's when I picked up R for the first time. And by the time I finished, I was proficient at R. And he happened to see a blog post by Andrew Gelman saying there was an R meetup that night. And so we're like, okay, let's go check out the R meetup. And this was maybe the fourth or fifth meetup that I'd ever been held for this group. And it was maybe 20 of us in a room up at. Actually, it's down at NYU at this point.

And what year was this?

This was 2009.

Right. So, yeah, this is. This is way back when.

Yes, yes, indeed. And it was this great room full of a bunch of people to go on to become great friends with, like Drew Conway, Harlan Harris, Josh. Re all of us in this room just being nerdy about this then obscure language. And so that was a great break for me. And then shortly thereafter, a few months later, another professor of mine from Columbia said, jared, there's a project I can't do. I need you to go to Myanmar and do it for me.

Right.

And having never done anything like this before, he talked me into this. And it was a great experience spending three weeks doing a humanitarian survey in Myanmar after the cyclone that tore through the country, and I believe 2008. And we were doing a periodic review of how they were recovering.

Interesting.

So after I did this, this is my first experience going and working on a data science project. They didn't call it that at the time. It was a stats project. That's what they were calling it at the time. And so then I came back from there and just kept going to the meetup again and again and happened to get set up with the right companies where I did some freelance work for until 2011. Drew Conway, who famously ran the meetup from its inception up until this point, decided he needed to focus on his family, finishing his PhD and starting his company. So he asked if I would take over the meetup.

And this. This was what is now known as the R Open Statistical Meetup.

Right. So he had a few months before he handed it over to me. He changed the name from the R Meetup to the Open Stats Programming Meetup. And then right at the same time happened to be again, right place, right time. A friend of mine who ran the machine learning meetup, Paul Dix, asked me on behalf of his publishing company if I wanted to write a book about R. So the two things happened Just at the right time. Taking over the meetup and being asked to write a book.

Fantastic. And what was this book?

This is what became R for everyone.

Beautiful. And we'll include a link to that book in the show notes.

Nice.

Could you give us a brief rundown of what the book is?

I envision the book as the the way I wish I had learned R the first time around. I really want it to be. Here's how you get started. From installing it to how the command line works, to using RStudio all up through using variables and reading and data, doing data manipulation and plotting, doing list operations, working the way up into statistics and doing modeling and GLMs and then nonlinear models of machine learning or onto writing reproducible reports in shiny. It's supposed to be the whole stack in the order and the way that I wanted to learn in the first place.

Fantastic. And so you wrote this book. You're running the meetup at this point, but now, famously, you run a large conference in New York City. You also have your own business, so maybe you could tell us a bit about these aspects of your work.

So the conference grew out of the meetup. This meetup is a special place to me. I really see it as a place where all these different people can all come together and be comfortable in a room. And I say this because a lot of people in this community are often maybe shy, introverted, didn't always feel the best in a big room of people. But you come here and it's just this loving, welcoming place that everyone feels at home. And I thought that, well, we do this once a month for a few hours a month. Let's have a two day gathering of just nonstop R and good vibes.

Exactly.

And it's this really special extension where. And that's the thing I keep saying. We go to this meetup and everyone just has a big smile and they're happy and everyone can be comfortable. And it's so important to me that that's how this continues to go as this great open space. And that's something. After our most recent conference, Jennifer Hill sent me an email saying that it's a room full of incredibly smart and incredibly welcoming people. And that's such a great combination.

Okay, so I agree completely. And I want to hear a bit more about the mission behind the conference and what particularly you're interested in. But the fact that Jennifer said it's a room I think is really important. I want to focus on that for one second. The fact that your conference is in one room Both days are in one room. It's single track means that people are essentially their hands are forced to interact with people as opposed to running around between different rooms. And you get a very different vibe by having everyone in the same space for that extended amount of time, right?

Absolutely. And that's part of it. We make it so that each talk's only 20 minutes with no questions. And after every three talks there is a 30 minute break. So you have to get up out of your seat and all the snacks are on tables right behind your seats. They have to go and talk to each other and meet the speakers in person and just be friendly and hang out with each other and make new friends.

And what's the idea behind the conference? I want to get to, you know, do you want people there who have a certain level of experience or would somebody who'd never programmed in R beforehand get something out of it as well?

I want it to be something for everyone so the novices can get there and learn all this great new stuff and the experienced people can be there and share the knowledge and learn from other experienced people.

Yeah. Great. So it's really the same philosophy that, that informed the meetup that you've been running all these years?

Absolutely.

Okay, great. So I want to talk more about, you know, these aspects of community building, but first I'd just like to have a quick rundown of the business you operate as well.

So the business actually started going back to grad school. I graduated in 2009 right in the depths of the economic recession and I realized that I didn't apply for a job. So what do you do when you don't have a job? You start a company.

Exactly.

And I started in 2009 and I officially incorporated a couple years later. And it is a data science. And if I have to use a buzzword, I'll even say AI if I need to use a buzzword.

Woo.

So it covers AI, data science, machine learning. I can rattle off a few more buzzwords if I need big data, of course, if that's the thing. And we are a training and consulting and advisory firm and we have multiple parts to this business. As I mentioned. First there's training. We go into companies and help educate them how to do data science. We tell them in specific how to use R, how to use Python, how to use SQL, how to use Stan, and we teach them how to use these tools in a very hands on manner, very much like my class at Columbia. And then on the consulting and services side, it can range from giving advice and strategy. It could be writing a white paper after a thorough analysis, or it could be about building a process and algorithm that's an ongoing mechanism that they use to run their business.

Great. And what type of businesses? I'm sure, you know, you run the entire gamut of verticals and industries, but what type of businesses do you get most requests from or work with the most?

So since we're in New York, there's obviously a large contingent of the financial firms because they are a dominant player in the city. And we also see a lot of pharmaceuticals and manufacturing.

Hmm, interesting.

Most famously because the one we probably got the most publicity for was we did the draft picks for the Minnesota Vikings.

All right, when did you do that?

This was the 2015 draft, the year that they picked Stefon Diggs.

Incredible. In terms of, you know, these types of industries. Have you seen these. These industries and verticals represented more at the meetup as well? Or, you know, have you seen the verticals change over the years of running this meetup?

Well, the meetup is so large, we're closing in on 10,000 members that the composition from each meetup really changes depending on the topic.

Yeah, that makes sense. So if you had someone who was big in, you know, the finance community, you'd definitely get a lot more people from finance coming.

Yep. And then when we have a more pharmaceutical talk, you see all the players from that field come. So it's really interesting seeing there's always a. There's always that mainstay core group that come to every meetup, but then with changing topics, you get vastly different groups. For instance, one time we had a talk about race cars, and there's a whole bunch of people who came to the meetup because they're interested in Formula one racing.

That's awesome. We'll jump right back into our interview with Jared Lander after a short segment.

Foreign.

It's now time for a segment called Data Science Best Practices. I'm here with Eric Ma again, a data scientist and research investigator at Novartis who helped to create our Python network analysis curriculum here at Datacamp. Last time you were here, Eric, we spoke about refactoring your code and writing functions. Today, once again, we'll import software engineering Best Practices as Data Science Best Practices with. With respect to testing. Eric, how do you make sure that your functions do what you want them to do?

Hugo, this is the world, the wonderful, wonderful realm of testing. Something I've come to realize as being super important for Data science, for testing. The basic idea here is that Once I've written a function that does exactly what I want it to do, I write a test for that function. Now, most good software engineers will actually do it in the other way, which is learning will first document the intent, then write the test, and then only then write the function. But for data scientists, oftentimes what I want to do doesn't become apparently clear until I've actually looked at the data and done some prototyping. So the tricky part here then is making sure that I write my test as soon as the function is correct and stable. But testing goes just, you know, it goes beyond just code. I think it's super important to write sanity checks for my data. So, for example, do I have the expected set of columns in the data set, or do I have an unreasonable number of null values? Or do the ranges of my data even make sense? Do I have negative numbers for columns that are supposed to represent counts?

I see where you're going, but it seems like tests for data are a bit hard to automate, aren't they?

It sort of depends. So for static properties of a data set, such as the columns, the number of rows, and the valid range of values, these are what I would call the schema of the data set set. And it's fairly easy to write automatically executable tests. Here, for example, it's not hard to automatically test that counts of physical things, should only have integer values and should never be negative. Now, the utility of these tests are to help me to check for data integrity and therefore catch errors when I accidentally tamper with the data set. On the other hand, for statistical properties of the data set, you're right, these are harder to test automatically. But more generally speaking, nothing beats visualizations. For example, I love using Justin Boyce's favorite the empirical cumulative distributions to visually check for potential outliers, which I can follow up on. I can also use tools such as MissingNo, you know, MissingNo, which can give me a visual overview of the number of null values in the data set.

Nice stuff, Eric. What's your favorite testing framework, by the way?

I like pytest for the following reason. It's very easy for beginners to get started, yet it's also super powerful for advanced users to customize. It really bridges well, the idiom, which is write a test function was really easy to catch on when I first started. And when I was more comfortable with the package, it was not hard to pick up more advanced features such as selectively running subsets of tests.

Yeah, Pytest is a great package to Pick up. And I actually remember helping you to teach a tutorial in which you use PyTest at PyCon last year.

That's right. It was really fun co teaching it with you.

Absolutely. And we'll link to those materials in the show notes.

Yeah, definitely.

All right, thanks, Eric. And see you back here soon for some more data science best practices.

I can't wait to get back here to tell you all about version control.

Hugo.

After that interlude, it's time to jump back into our chat with Jared Lander. So this really brings me to kind of the. The focus, the meat of this conversation. We're here to talk about community building in data science, which, as we know, you've been active in for, you know, over a decade in New York City. So I'm just wondering how, like, in all generality, you think about building communities in data science.

So the very first step is being welcoming. As a few people know by now, I unknowingly met my wife at the meetup. And I didn't realize it at the time. It took a year later until we met another way. And turns out we met a few ways until I finally got the hint. But she said to me later that one of the things she remembered is that I went right up to her, said hello, the big smile said, welcome to the meetup. I hope you have a fun time. And I don't just do it for my future wife. I try to do that to everybody. And I encourage all my members to say hello to all of the new people they haven't seen and make everyone feel welcome, because everyone who walks in the door, they all think that they know the least out of everyone in.

That room, particularly in a field like data science, I think.

Yes.

Yeah. So everyone feels insecure. They could be the smartest person in that room and they still feel like they're at the bottom. So whether that's true or not, and usually it's not, because everyone there is incredibly smart. Who cares? Come in, have a fun time, and nerd out a little bit and enjoy everyone's company.

Yeah, I think that's really important. Do you experience that? I mean, what I'm trying to say is that people who are extroverts already may benefit more from these types of situations. So I'm wondering if there are any mechanisms or life hacks to help introverts or people less willing to talk become involved in the community as well.

So we start every meetup with pizza and beer. And everyone loves pizza. And beer is usually liked by a lot of People. And it just starts already. It starts, it starts with a convivial environment at that point, because everyone's. You're eating, you're having some beers, you're standing around chatting. So it helps the people who might be less comfortable to get into a groove. And then when we all sit down, we all take our seats. The very first question I ask, and this isn't a social question, but it's who's hiring? And already it makes it an open environment. Hey, who's looking for, who's looking to hire people? How can we get people jobs? It makes it a welcoming, friendly place. Then the talk happens. And then afterwards, I encourage everyone to go to the local bar, whether they drink or not, they can hang out and be friendly in a place where they now know that everyone else there is just like them. Another data nerd.

Yeah. And something that you always, in my experience, have asked people to do or almost stated people have to do when they stand up and say they're hiring for certain positions is you say, go and stand up the back afterwards, and anyone who wants to go and chat with you, go and do so. So you're quite passionate about getting people to connect in, in that way and forcing people to talk to each other. Forcing in inverted commas, of course.

Yes, yes, indeed. Because sometimes people just need that little push and then they have everything in the world to say.

And how about, are there any other ways you think that's important about building community building in data science?

Yeah, because not only are data scientists in general going to be reluctant to put themselves out there, which could be a security thing, it could be a humility thing, but also, you see, certain groups within the community are even less represented and even less likely to talk. So I want everyone to be welcome and find a home here. And it really does feel like a home. So you need to make sure everyone feels welcome. I want everyone to feel at home. And it really does feel like a home. And that means everyone, including underrepresented groups.

And how do you think about creating that home for underrepresented groups?

So the first step is treating everyone the same. Make sure that everyone who walks in the door gets a handshake, gets a smile, gets a hello too, regardless of who they are. If they're an old friend, someone new, someone different, it doesn't matter. Everyone needs to be treated warmly and with welcome to reach out to underrepresented groups. I find it's often good to have what I'll say is like a lieutenant, we've done a really terrific job lately of getting parity at the conference between men and women. And a large part of that is our ladies. And a number of my friends are at the top of our ladies. They say, hey, who should I have speak? Who should come? Can you tweet about this? And that's very helpful. And then Emily Robinson, who now works at datacamp, came up to me, and she might not have worked there when she gave me this idea. Actually, I think this might have been before she took the job saying that, what if we offer discount codes to underrepresented groups? Because now that we've hit 50, 50 men and women, let's offer this discount code. So she tweeted out, hey, if you're an underrepresented group, message me. And we have discount codes for you. And it's a way to get people to come, and we cut the price significantly. That's an important part about both the conference and the meetup. We never let price be a factor in people attending, whether it's the $5 admissions to the meetup or the conference fee. If you can't afford it, we'll get you in there somehow.

Yeah. And it's a practical measure, but it also sends an important message as well.

Absolutely. Money should not be a barrier to learning and feeling welcome. Neither should any other circumstance you are in.

And you mentioned having a lieutenant. Can you speak a bit more to this and what this means?

Yeah. So this community, as I said, they are so welcoming and engaging and warm. Everyone wants to do something, and everyone wants to chip in somehow. So I see it as my way of saying, you want to contribute chip in. Yeah. Here's what we could do. So whether it's Emily giving out discount codes, underrepresented groups, or Sumya blasting it out to the R ladies, or the people saying they want to draw paintings of the meetup, or people wanting to sponsor it because they want their company involved, any way that I can help people engage and then they in turn engage more people is a win.

Yeah, very much so. And that really speaks to the idea of, you know, drawing paintings, for example, isn't something I would have thought of, but it speaks to the idea of keeping a really open mind about avenues of getting. Getting people involved and getting as many groups involved as possible.

Absolutely. In fact, at the conference, Thomas Levine, who has been a longtime member, he famously gave a meetup about how to make music videos in R. For the conference, he hand painted data visualizations of famous data sets, and we auctioned them off to support the free Software foundation and the R Foundation.

That's fantastic.

Yeah, it's so much fun.

This actually raises another interesting point. We haven't really talked about a lot about content in conferences and or meetups. We've talked about kind of the social aspects and how to get people. People involved. But how about consistently providing good content? I mean, I think about this a lot at the moment in terms of the podcast, for example. But yeah, how do you think about providing good content and diverse content over the course of months, weeks and years?

Eventually the content is really tough, but it's key because people are coming for good knowledge. I heard someone once say about the meetup, he was standing behind me. He said to someone else that this is the best night school education you can get, and it's free and you get pizza. Exactly. You get pizza out of it. I mean, what more do you want? So I spent a lot of time chasing speakers. Now, some speakers, I say, hey, can you speak next week? And they're like, sure. Other speakers, I have to ask them a year in advance. And I'm constantly finding speakers from a vast array of topics. And I try to be cognizant of our 10,000 members. They're all different skill sets. So one month we'll have a deep reinforcement learning talk, and the next month we'll have how to use dplyr Key is whether it's an advanced talk or a beginner talk, is having a good speaker deliver it, someone who can talk to the crowd and be engaging. And that comes from me knowing my members, knowing who they are, talking to them in person, seeing how good they can speak to people and how well they can present. But it also involves outreach. Every year I send a letter to John Chambers, the person who created S. Do you want to speak this year? Every year I reach out, and it has worked for me in the past. I've gotten speakers. Not John Chambers yet, but for instance, Rob Hindman, the famous Australian who wrote the Forecast package, is coming to New York next month to speak at both the meetup and to give a workshop for us.

Fantastic.

And that took. I emailed him over the course of about two years, seeing when I can get him to come to New York, and eventually it lined up.

Yeah, that's really cool. And similarly, you know, we've been chatting about you coming on the podcast for some time, and here we are.

Exactly.

I think the point is you've got to. You got to do the time, right? And you. And you've got to do the work. And essentially, you know, I Don't want to, you know. Well, I will use the word. You got to bug people as well, right?

Yes, yes. You need to know how to push the buttons on the person to get them to both want to speak and be comfortable speaking and put them in a good place to do a good job.

And so providing good content in this framework is one thing, but also providing content that's essentially super important for either beginners or practicing data scientists. Right. So deep reinforcement learning speaks to something that's really hot and at the moment and getting increasingly more important as is. I mean, a couple of months ago you had JJ Allaire of R Studio talk about Keras in R Now, which is something that's very pressing and very important. So you've got to keep your finger on the pulse of the entire community and the field, right?

Absolutely. You got to know what the cutting edge stuff is, what everyone's excited about, and what's going to be the cool new things to help them do a better job.

And how do you do that? I assume having already developed such a large network of not only colleagues in data science, but you're also friends with a lot of people in the community helps, but what other techniques do you have for really keeping your finger on the pulse of what's happening now and what will happen in the future?

I just had to voraciously keep reading everything that's out there, whether that's our bloggers, Andrew Gilman.com, our views, even things on ZDNet and CNet and different computer magazines and Lifehacker, just keeping abreast of everything that could be touching our world and really almost doing a deep dive in all of that to see what's going on. And that's both for the meetup and for my business, so I can keep aware of the current trends and what we need to do.

And is this a profession for you or is it really a way of life?

I want to say it's absolutely both, because I love what I do, I'm doing this, and if I'm not working on a client, I'm doing something for R in my spare time anyway. I'm just constantly clocking away doing data science, whether I'm getting paid or not.

Yeah, absolutely.

It's all encompassing.

So this has been a great tour through your approach to building communities in data science. I'm wondering if you can speak to some of the challenges you found in community building in data science.

Well, the first one, and this is a good problem to have, is underestimating how popular A speaker is going to be. I know that sounds like, you know, it sounds like a humble brag. In reality, it's my job to make sure there's enough space for everybody. And I hate when an event completely sells out because that means there's people who can't come.

And you don't want to be consistently a meetup where people can't get in. Right. And people get frustrated with that.

Absolutely not. Because we wanted it to be open to everyone. So we have started lately live streaming the meetup.

That's cool.

And that's because a, we realized we have members all around the world who want to come. We have membership in the Netherlands and Australia and Israel and Singapore and London, you name it. We have people all over the place. So. But also importantly for if people can't come or they want to watch it later, we started live streaming in putting the videos up online, all free and available, of course. So that is something that I find to be very important. Just another way to make it accessible.

What other challenges have you found?

Another big challenge is, and it's sort of what we touched upon before, but making sure it is welcoming for everybody. You don't want someone coming in the door and not being greeted. I don't mean necessarily literally greeted, though we try. But someone coming in the door and not feeling at home there. And that's something that we really have to work on because not everyone wants you coming up and shaking their hand and talking very loud. Some people want to be quiet and maybe want more of a one on one experience. And so that's a very big deal. Finding a way to make everyone feel welcome in a way that's comfortable for them.

Yeah. And I suppose, are there any challenges you found that are particular to New York City that you don't think may be experienced elsewhere?

Well, it's definitely a space challenge. Space is at a high premium in New York City and it's very difficult to find a big enough space for enough people on a regular cadence to walk in the door and have a good time.

Yeah. Something I have noticed though is you've strategically made partnerships with certain organizations. I mean, I went and I went to one of your meetups once, which it was actually Hadley Wickham talking at the Twitter headquarters in Chelsea, if I remember correctly.

Yes, indeed. That's a large part of, I guess, my life. Like you said, this isn't really a career. It's a all encompassing calling for me. It's all about being friends of everyone and it's all about developing the personal relationships. I try to develop a relationship with everybody. And then because of those relationships, I happen to be able to have places like Twitter is available for the meetups. They realize how important this community is. It's all about being nice and friendly to everybody. And I try to do as many favors as I can to people because that's what a good person should do.

Yeah. And I actually remember one of the first times you and I met, you were like, hey, man, at the conference, I'm going out to dinner with some people. Come along to dinner. And now whenever I see you at a conference, we go and have drinks and dinner and all of that stuff. And you're always inviting everyone along, which is a wonderful quality, I think.

Thank you very much. Yeah, I think that's very important because I remember when I first started going to this group, there's 20 of us and we were all friendly. And I, like many people in our field, went through an awkward phase definitely at some point, and it would have been great if someone invited me to dinner. So now everyone should be welcome at dinner and everywhere.

We'll jump right back into our interview with Jarrett after a short segment. Now we'll take a moment to check in again with Peter Bull, co founder at Driven Data, and see what's new on our segment, Data Framed for Social Good. Peter, thanks for coming back to share more about the Data for Good movement and how to get involved.

Hugo, you know, I can't resist the opportunity to chat data with a bearded Australian. For me, that chance does not come very often, so I need to seize it when it does.

Well, I wouldn't want to deny you that, Peter. Last time you gave us some historical insight into the origins of data for good. What's going on today in the Data for good space?

There are so, so many ways to get involved, and it's my belief that everyone who is a practicing or aspiring data scientist can find a meaningful way to contribute. There really are few things as gratifying as using your unique professional skills to make a difference.

We spend a lot of time on the podcast talking to data scientists about where they work, and I've had the pleasure of great conversations with folks at tech companies, in traditional industry, at consulting firms, and in academia. Lots of these folks are using data in ways that have an impact. Epidemiologists, biologists, people working on fake news detection or open source projects. To me, this seems like data for good.

Right?

That's a great point. Data science is a toolkit for problem solving. So the for good aspect is Largely a matter of what problems you choose to work on, not any particular method. In fact, I believe that data for Good encompasses a large range of projects and that we're a welcoming community for many different kinds of contributions.

So if I want to do data for Good, I should go work as a data scientist at a non profit or ngo.

Let's talk about that for a moment. Do you know the median salary shout out to being robust to outliers of a data scientist in the United States? And I know the reach of this podcast goes far beyond that, but sometimes we must work with the data we are given.

Well, we've seen the huge demand for these skills, which is why we're trying to teach them at data camp. I imagine it's easily north of 100,000 a year.

That's right, Hugo. Or as you might say, you eat the bongo on the barbie.

I don't think anyone would say that.

Oh, well, maybe my Australian's a little rusty these days. The median salary is around $120,000 a year, and demand for data scientists, like all good charts, is going up and to the right. Now, do you know the median salary of the executive director of a medium sized US Nonprofit?

I think a core skill of a data scientist is being able to admit when you don't know something. And I'll be honest, I don't know.

That really is one of the core skills. And the median salary is about $130,000. And that's the top of the pyramid for these organizations. So I'll let you figure out the regression for about how many data scientists these organizations can hire.

Oh, Peter, my calculations are coming in at less than 1.

That's about right. And the point is that if we want to use the power of data to help people, we need to think beyond our jobs. If people want to get involved, there are volunteer organizations, meetups and hackathons, fellowship programs, project marketplaces, digital data for good communities, social impact data sets for researchers, and near and dear to my heart data for good machine learning competitions.

I'm sold. How and where can I get involved?

Well, Hugo, I'm very excited to be talking with you about that in our next couple segments.

The suspense is going to kill me. Peter, time to get straight back into our chat with Jared. I think the one thing that we've been circling around that we haven't mentioned explicitly is we need to remember that people are taking people work full time, right? And they're taking time out of their evenings, away from their families, whatever it may be to come along to these meetups. And so we need to be respectful of that fact as well.

Yeah. So you need to make sure it's worth their time coming because it is a commitment, especially since we hold these in Manhattan and while a large part of our group are New York City based, but there are people coming from New Jersey, Connecticut, Long island, and it's a long way home for them, so it really better be worthwhile for them.

That's a great point. And I actually remember the first meetup of yours I came to. I was living. This is a bit of a personal journey now. I was living in New Haven, Connecticut, and I got the Metro north down, which, you know, door to door took me. Took me two hours. And I was working basic science research, wanting to get a feel for what, you know, industry based data science look like in, in practice. And I arrived at your meetup and you said straight away, everyone who's hiring, get up and talk about it. And there were people in finance, in tech, in, in health, in management consulting getting up. And it was my first kind of introduction into this wonderful world of, of data science in, in the city. And, you know, I got back to New Haven and then cycled home and got home at, you know, 1:00am or whatever it may be, but was actually really invigorated by this community and found it really inspiring.

That is fantastic.

And now I work at Data Camp.

Look at that.

So here's my version of a success story. I love it in general in terms of community building for anyone in any city, in any country. What are the top three things you would encourage them to do?

So, of course, beyond finding a space, which is crucially important, you need to find the speakers. You need to get these speakers. And it can't be the same speakers. It can't be the same one every time. You need to have variety in there. I let my speakers repeat after a few months, preferably a year in between. One person speaking. We need to have this variety of speakers and topics. Now that's for the organizers, but for everyone who attends, step out of your comfort zone a little bit. I know I keep saying this on this podcast, but say hi to everybody. You never know who you might meet. It might be somebody you could collaborate with, someone you just want to hang out with, someone you know through Twitter. I've been at definitely these events and I meet someone and say, oh, you're this handle from Twitter and that's how we get to know each other. But for other people who also want to get more involved, they should write about what they do, write a blog post, write about the analysis they did, talk about what they're doing, and then give that talk to the group. Us as organizers are always looking for people to speak. And I love it when people come up to me volunteering, hey, can I give a talk on this topic? And I'll sit with them. They've never done a talk before. I'll sit with them and make sure it's going to be a good talk. And that's actually important from an organizer. Support your speakers, help them craft the talk that will reach your audience well, but then for the people, give a talk. Get out there and share your knowledge with everybody else.

Yeah, I think that's really important. And I like the idea of an organizer collaborating, maybe not being too hands on, but, you know, being a board for bouncing ideas off with respect to the talks that happen, as opposed to just saying, you do this, come and.

Give it, and it really benefits both the speaker and the audience. This way you make sure the speaker feels good about what they're doing and they do a good job. And then for the audience, you're making sure that they get a good talk.

The other thing is, I always encourage people to say hi to everybody at meetups and conferences and. And these types of things. I do have a concern that for people like you and me, well, it's easier said than done. I mean, people like you and me, generally, our natural state of existence is saying hi to people, essentially, right? Like, we're, We're. We're inherently very comfortable doing that. So it isn't so much a stretch out of our comfort zone. And I wonder if there are any hacks for getting people started. I suppose your idea of getting people to stand up and talk about positions they're hiring for. My El Salmon, who I had on the podcast earlier this year, told me about a conference where there's a buddy system set up whereby a newcomer would be assigned to you or Emily Robinson or Hadley Wickham or whoever it may be to hang out with them for a bit, to get a feel for the community in that way. So I think there are clever tricks.

Like that, and I think you said the word in there, force. And I have at times tried to force people to put themselves out there. And I think two of my favorite anecdotes about that is, especially amongst my own students, and I see they're very shy, and I know that they're really good. And so I'll be like, hey, you're speaking at this conference here's your time slot. I don't really give them a chance. Not asking if they're going to speak. I say when they're going to speak.

That's it. I remember Emily actually telling me that you'd emailed her and say, hey, congratulations on being chosen to speak at my conference. Let's sort out the timing.

Exactly.

I love it.

Yeah. And that's. Some people just need that little push. They need to. They need that, that little push to get going and that's all they need. And then they go off and running on their own. And another thing, I think Hillary Parker famously said this in her podcast that I've been trying to get her to speak. So then one day I just put her up on the website that she was speaking.

That's brilliant. And did she?

Yeah, absolutely. With gusto.

I actually remember just before the election, you had a meet up in which Dave Robinson and Hillary both spoke kind of themed around. Well, Hillary spoke about the popularity of the name Hillary, or lack of popularity, and Dave Robinson spoke about his analysis of the Trump tweets. Is this common to get a couple of speakers involved at the same time? Because that was really great at generating a lot of buzz and social activity.

I do try if there's a common topic to put speakers together, but that could also be very difficult, coordinating speaker schedules and venue schedules. So if it comes together, I love it when I can get them together. But though that's even harder to do than you would imagine.

So how does your consultancy work and your business play into this conversation about community building?

As you've seen, as you know, the community is very important to me. So for the conference, it's a little more obvious that it's a conference brought to you by Lander Analytics. For the meetup, however, my company is more of a silence sponsor. I think it's very important that the meetup and my company are separated in that my company is not seen as benefiting at all from the meetup. It's just this meetup is a special entity to me. It's this precious thing that is open source and for the community. That said, the meetup is a lot of work and I use the resources of my company to help run the meetup, whether that means paying for things out of my company's budget or means having my employees do various things for the meetup, whether one person's running the camera, one person is checking people into the door, someone else is making sure the food is in place and the podium is set up. So the company sort of invisibly helps things run in the meetup to make sure it's going because it's so important that that meetup is this non commercial space.

And you've actually mentioned something really important in there, which I don't think we've really touched upon, is the fact that this isn't a one man show. You have a company of incredible colleagues, you know, working on this with you, right?

Absolutely. It takes these meet have now grown so big and so complex. It takes a lot of hands working on it. And that's one thing actually that I've been very fortunate because I managed to build a company to help with this. I went to a conference for meetup organizers. It's a little meta, but it was a nice gathering at the National Science Foundation, I believe. And one of the biggest concerns for the meetup organizers and most of their meetups weren't quite as large as mine, but their biggest problem was funding and time to do all this. Now, while I still am the driving force behind the meetup, I've been fortunate enough that I could dedicate my own company's resources, both personnel and money into helping make this thing a little bit better and drive it, whereas these other meetups don't have that luxury.

Yeah, it's interesting that you mentioned going to a conference for meetup organizers because I actually, I'm thinking of going to some conferences for people who make podcasts.

That's awesome.

Very soon. Which is never something I thought I would. I'd think, let alone, let alone say. But it's, it's pretty exciting because we've been talking essentially about community building in data science. We haven't talked a lot about technical data sciencey techniques, but something I ask a lot of my guests and something I'd like to know from you is what is one of your favorite data sciencey techniques or methodologies?

So my answer to this actually ties in nicely to data camp because I think I have the dubious distinction of having the longest outstanding class due to you. I've owed this class to you since about 2014, back when the company was just in Belgium.

Fantastic. And it may not even have been called datacamp.

No, it was at that time, I think. Actually you know what? It might not have been. You're right.

Yeah.

So I owe this and.

Great. Well, I can't wait to see this course. What is it?

It's about glimnet.

So tell me about glimnet.

Glmnet is probably my favorite machine learning data science algorithm model. It is the elastic net which is a dynamic combination of both ridge and lasso regression. The reason I love it so much is that it does automated variable selection and shrinkage for a linear model. It's amazing. And you could throw thousands of variables at this and which would ordinarily give you multicollinearity and overfitting, and it reduces the dimensionality down to a point where you have a strong stable fit. And at the end though, you still have a linear model, so it's still interpretable.

That's really cool. And so I think for those more technical out there, ridge and lasso regression are forms of regularizing your linear model, putting more constraints on the choice of parameters, right?

Yes, 100%.

Of course, you know, this is a great technique, great tool, But I think also one of the great things about glimnet is that the API is so easy to use as well, right?

Oh, it's so amazing. You go do with one line of code in R, you give it an input X matrix and output Y vector, and of all the defaults, that's all you need to do these two arguments. And then if you want to do cross validation over the lambda hyperparameter, you just change it from glimnet to CV glimnet and it takes care of it all for you.

Well, I for one am super looking forward to this course for many reasons, but just the way you talk about it, you sound really passionate about it. So that's super exciting.

It's a great model algorithm and I love this thing so much. It's blazing fast and it's my favorite part. It's written in just 73 lines of Fortran code.

That's amazing. I wasn't aware of that. That's really cool.

Right?

So we also, we've mentioned that to organize your conferences, to organize your meetups, to use these types of structures to build communities, you need to know what's happening in the data science community and even to know what's going to happen in the future. So my question for you is, what does the future of data science look like to you?

So traditionally in data science we have a bit of a divide between having strong predictions and having interpretability. And depending what task you're doing, you have to sacrifice one for the other. But I'm seeing more and more advancements in interpreting strong predictive models. Whether that's putting confidence intervals around glimnet coefficients or using Bayesian additive regression trees to get confidence regions for causal effects. I think we're starting to see ever put into doing inference around high predictive models And I think the future will be having a very strong model and being able to explain why it's happening.

Yeah. And we're seeing a huge push from many stakeholders, whether it be customers, users of products, legislation and businesses want to know why their models say what they, what they do. Right. They want to be able to understand them in some sense.

Right. You don't want to just turn over the keys to the black box because then you can hide what you're putting in there. So people want to see what is going in and why and how it's doing it.

And that's one of the wonderful things about, you know, these types of regression we've been discussing and logistic regression, of course, is that you can say to someone non technical in this model, if you increase this parameter, that amount that results in this in the output.

Exactly. And that's what makes these linear models so attractive.

So my last question for you is for all our listeners out there who are interested in data science, who are interested in being part of communities, what's a final call to action for all of them?

They have to show up. The first step is walking into that door, sitting down and watching an amazing talk, then sharing a slice of pizza and a beer of your colleagues and getting involved.

That's fantastic. So everyone out there, to reiterate what Jared said, you have to show up. And I couldn't agree more. Jared, it's been an absolute pleasure having you on the show.

It's been equally for me.

Thanks for joining. Our conversation with Jared about building communities in data science. We saw that a key aspect of community building in data science is welcoming everybody at the door, providing a convivial and social environment, not just the dry information flow and getting everyone talking about all types of things by, for example, getting everybody with job openings up and talking about them. At the very start, we also saw that it's of the utmost importance to let people know that we all have insecurities about our practices, whether we're just getting started with data or have been data nerds for years. Jared also made clear the importance of making the community feel like a home for everybody, including underrepresented groups, and cited specific measures he takes to this end, such as having discount codes for underrepresented groups. Make sure to check out our next episode. A conversation with Jenny Bryan about the use, misuse, abuse, necessariness and future of spreadsheets in data science and data analytics. Why are spreadsheets so ubiquitous in data analytics? Why are so many data scientists anti spreadsheet join us to discover why spreadsheets are in fact necessary in data analytics and how spreadsheet workflows can be incorporated into more general data science flows in sustainable and healthy ways. I'm your host, Hugo Bown Andersen. You can follow me on Twitter Hugo Bowne and datacamp. At datacamp you can find all our episodes and show notes@datacamp.com Community podcast.

It.