image: thumbnail.jpg
Data Science Hangout | Regis James, Regeneron | Achieving scalability & showing value of community
Achieving scalability through standardization (19:04)
In order to achieve scalability I do have to standardize things a bit. We use Confluence, so on the wiki that I've made, I have a lockdown page where I have some standardized messages that I send out to people at different phases of things.
I do customize and read what people say about themselves and I'm thinking, “this could be useful here, that could be useful there.”
One of the things that I do is– I have this whole ingestion pipeline that I maintain to identify new members of the community and who can help the community learn more about the blockers to current responsibilities they have for their job.
All companies have a directory where you can learn about who exists at the company and see people's job titles.
I can look into the directory for data or data science or whatever and see if the people that come out in the results are already in my Microsoft Teams group because they could be potential new people who could benefit from being in the community.
They might be able to do their job better and also help other people do their job better, if they can connect to others who have similar needs or can offer similar things.
I do this on a semi-regular basis, so I can really help onboard people who have just joined the company or maybe moved into a data science type role.
Then I reach out to these people, and use my standardized message like, “Hey, I just want you to know this group exists. This community was built just for people like you. This is what we do and here's a link to the wiki that has the links of all the recordings” - because we have a growing library of all the conversations just like the Hangouts.
I only send one email and ask people to let me know if they’re interested in it. I don't spam people.
Of the subset, who answer - sometimes it's like 20% that respond, they say, “Yes, this is amazing. Thank you for letting me know. I just came to the company. I didn't even know if there was a community.” I'm also doing it from my own perspective of when I joined the company.
When they respond and then they say they're interested, I add them to the Teams group, to the Active Directory group, and then forward the invitation to the ongoing gatherings every other Friday.
In the invitation text I paste in, “Welcome to the group. Please introduce yourself in the introduction section" and include a link that takes them directly into not just the Team for the group, but the channel of introductions.
Since doing this, I've gotten a lot higher percentage of people who have been added to the group to start introducing themselves and say what they work on.
I see what people are working on, and integrate that with my understanding of what people need help with from the Shiny app (the poll for what talks people are interested in.)
Then I reach out to them and say, “Hey, would you be interested in being one of the guests for an upcoming event?” I send them the link and they can see there's no pressure because I'm a year out. I've got someone every other week. So it's like, Hey, would this be something you’re interested in and then I am gently having a conversation over time.
Sometimes people drop out and I just reach out to those people and ask if they would be able to switch.
______
How do you show the value of community? (40:14)
To prove the contribution, the benefit of a community – I use the words stuck and unstuck because that's really the only thing the business, or any organization cares about.
There's a bunch of stuff that the organization has to do. There are obstacles in the way, they are stuck. How do they get unstuck?
That's what data science is for, really. You can argue that it’s also for going from stuck to unstuck.
The non-computational people who have the power to bestow upon you promotions, more money, or more influence over the company to make other decisions, really care about the value of the institution.
If you can show how the existence of that community propagates through the network that exists in the institution so that something goes from stuck to unstuck and you can mark that, you can measure it, then you can say this only happened because of me.
If you're a community that's gathered around pulling thorns out of your colleagues arms, you can say, we pulled 15 thorns out or we taught another group how to pull 100 thorns out because we were force multipliers.
If you say, “what do you hate about your job?" to the person with power that I need to prove the community's value to, think about how you’re reducing those pain points for those people and enabling them to go from stuck to unstuck.
Collect those things and showcase them. From the get go, you should be thinking about how to indicate success to leadership.
To join future hangouts: pos.it/dsh (All are welcome! We'd love to see you!)
Transcript#
This transcript was generated automatically and may contain errors.
So happy Thursday, everybody, welcome to the Data Science Hangout. Hope everybody's having a great week. I'm Rachel. If we haven't met yet, let us know if it's your first time joining, you want to say hello in the chat so we can all welcome you in together.
This is an open space to chat about data science leadership, questions you're facing, and getting to hear about what's going on in the world of data across different industries. And so every week, we feature a different data science leader as my co host to help us lead the discussion and answer questions from you all. Together we're all dedicated to making this a welcoming environment for everybody. And I love to hear from everyone, no matter your level of experience or area of work. It is totally okay just to listen in if you want.
There's always three ways that you can ask questions, though, and also provide your own perspective on certain topics. So you can always jump in by raising your hand on Zoom. You can put questions in the Zoom chat. And feel free to just put like a little star next to it if you want me to read it if you're in a coffee shop or something. And then we also have a Slido link where you can ask questions anonymously. And Hannah will share that in the chat here in just a second again.
We do share the recordings of each session. So they'll be up on the Posit YouTube, as well as the data science Hangout site. But with all that, thank you so much, Regis, for joining us as our co host today. Regis James is a senior manager of biopharmaceutical data science at Regeneron Pharmaceuticals. And Regis, I'd love to have you introduce yourself and maybe start off by sharing a little bit about your role and something you like to do outside of work, too.
Regis's background and journey to Regeneron
Sure. Thanks for having me. Let's see a little bit about my role and what I like to do outside. So I've been on the side of the company that I'm on right now for almost nine months. But before that, and I'll kind of clarify what I mean by that, I came to Regeneron right out of my Ph.D. in 2016, and I was on the R&D side and I did a lot of collaborations with biologists who were generating tons of data, but it was kind of challenging to make decisions based on that data. So what I did a lot was distill things down to their fundamentals to simplify the process of going from stuck to unstuck.
And in June of last year, I moved over to be a bit closer to the patients, which has always been my passion ever since before I went to undergrad for bioengineering. But I wanted to be closer to the clinical trial side. So that's where I am now. And I'm working to help optimize clinical trials using AI and ML. But there are a lot of different parts of the side of the company that I've moved to, which is called global development, that are not just clinical trials. There are a whole bunch of other things. And we are in the process of figuring out how to establish some of these collaborations with them as well.
Building an internal data science community
I also have, since the time that I was on the R&D side, have been maintaining a community of data scientists that are around the entire company. And since Regeneron is global, it's actually a global organization. And the reason why this organization formed was because when I came right out of my PhD, the ecosystem that I used was R. And RStudio now posits environment of the RStudio IDE and Chinese Server Pro, which has now been evolved into RStudio Connect. So I needed to get that set up so that I could do the same level of collaborations with non-computational people that I did during my grad school work, but at Regeneron.
So when I set that up, other people heard about it and they were like, oh my gosh, this is amazing. Can we please do this? So then I started letting people on. But I don't want to say with great power comes great responsibility, but with great open source access comes great responsibility to support an ongoing community of people. So that ended up happening.
but with great open source access comes great responsibility to support an ongoing community of people.
And what people were doing was building Shiny apps and a whole bunch of different other types of analytics that they were sharing with non-computational colleagues. And so I started hosting these showcases that would show, I hadn't mentioned this part to you, Rachel, but that would show what people had been building. And then they would kind of show it to a range of people around the company, which was kind of, it was feeding, it was a positive feedback loop because people started understanding, as my current boss calls it, the art of possible. So then they were like, wait, that's, does that mean we can do this? And can we do that? Can we do that? So then it kind of started growing.
So then I was only working in the R side and supporting that community. But then similarly, I can't remember which one happened first, but it was very close in time to when RStudio was like, there's other people in the world who use data science. I mean, the IDE, the workbench always supported a broad range of languages, but to make it more formally cross language, that happened around the same time that I moved the group from just R people to R and other programming languages that exist to enable data science. So that's the group that I convene every other Friday.
And yeah, we're having one tomorrow. It's internal, of course, but we're, so it's, we talk about the problems that data scientists face and how we address them. And then it kind of often, especially this year, it's been triggering spinoff meetings of other people who have common problems and they exist across the company. So it's been going pretty well.
Personal interests outside of work
In my free time, I mean, I am a lover of sci-fi things. So I have for a couple of years been working on a data science novel series that takes all of the geeky, nerdy, awesome things like cryptocurrency and CRISPR and a whole bunch of different things and integrates them into what I think is going to be an interesting story. And I'm planning on releasing that at some point in the future. So that's what I do in my free time.
And I've been able to influence or I've been able to attract my wife to also contribute. And she's also a scientist. So sometimes we have these like brainstorming sessions about different plot points that could happen and making them realistic, but also having a bunch of action in there too. So that's the thing I do in my free time with my favorite person. So I love it.
Facilitating community spinoffs
It sounds like you have a lot of people who want to buy that in the chat here right now. So when that's out, I'll have to put it on the list of recommendations to all the books Libby's been collecting. But I'm so excited that you started off with community because that's something that I was going to ask you about as well, because I was leading a community building workshop at R&Pharma. And during that workshop, I personally learned so much from you Regis and just wanted to be able to share that with everybody here too. But you just said that a lot of people start having these these spinoff meetings from the Friday sessions, which is amazing. And I was just wondering, how do you help facilitate that within the community?
So they haven't always been like the spinoffs have kind of started a little bit more recently. Sometimes there have been a thing here or there. I haven't, to be honest, and everything's a work in progress. The community is about helping each other figure out how to figure stuff out. So in the most recent spinoff, I wasn't exactly sure where my role was. They were like, thank you, Regis. We are together. We're meeting each other because of the organization that you created. And I was like, yes, cool. But I'm still trying to figure out exactly what it is. And I mean, I don't think I should be in charge of everything. It's good for things to happen organically, but be triggered by existing problems that people have.
The thing that happened recently without because, of course, I can't go into detail about what it was, but I can speak in general, but yet specific for this community data science terms. It was about natural language processing in different contexts to identify things that we need to act on. And there are three different parts of the company that have the same problem. And we would not have come together without that. So people were asking questions.
What I typically do, though, when I prep the speaker. So it's every other Friday. And before last year, I was setting a meeting. Actually, I was doing it even more lightly. I was just messaging them on Microsoft Teams to say, hey, are you ready to have XYZ so that we can kind of share it with the community? But as things have gotten more complicated and people's schedules have gotten crazier now, I I have my schedule set out for the whole year, basically until the end of twenty twenty three every other Friday, except for holidays. So now what I've done is I even have I have the page open right here on my other screen. I have a column of red, green and yellow icons indicating whether or not the schedule date is locked down or not. So whenever I have it locked down, when the person agrees, then I schedule the two weeks before. So not the Monday of the gathering, but the Monday before I I meet with them to make sure we have everything together. So so that it's really easy for the people who are in attendance to know what the walkaway scenario, what the takeaway point is, so they can apply it to their own work as quickly as possible.
So I do a lot of honing there and I think it's been paying off because in the meetings, people have been really engaged and have been asking questions during the meetings because I've been able to whittle it down. The speaker that I have tomorrow, I said, I apologize. I don't mean to ask you to ruin your story, but it may be helpful if you'd say the end part at the beginning. And he's like, no, no, no, please, please. I do want to ruin it. So I think that would but but having them basically ruin their stories and saying the end at the beginning is what gets I think it helps people get engaged. So there were different departments that got engaged because of the way I was able to flip the narrative and say the ending from the beginning. And so because of that part, I think people were able to just jump in and identify things that were needed.
having them basically ruin their stories and saying the end at the beginning is what gets I think it helps people get engaged.
And so kind of it's a long winded way of saying that I was sort of sitting and listening to the subgroup from the community talk. I was involved in one of the projects that the that spun off that were discussed in the spinoff. But yeah, it was more just kind of sitting and listening and figuring out, oh, this is a new thing. How where do I sit? If anyone has recommendations on how I could probably engage in that, I'm all ears.
That's great. Pulling that up to the front so that people like know why they want to pay attention to it, too, is I was trying to grab the link. But I remember many give a pause. I gave a talk a few years ago about like letting them eat cake first when teaching students and like showing them the pretty application first that they might make you. I see you have your hand raised. You want to jump in?
Audience Q&A: NLP approaches
Sure. Hi, Regis. First, this is Yu Kang. I'm a data science professional from a finance company. And I have a technical question. You just mentioned that you are doing an ILP model. So could you share some details of this? Because we are doing something similar. So for your ILP model, what's your raw material? Is this a conversation over the phone or some transcript? What is that?
I cannot go into detail about it. Because it's proprietary company information, at least what the data is. But I can say at a higher level that one of the so one of the three chunks of the company that was represented and multiple people from each of those. Well, one chunk had only one person, but another chunk had two or three people and another chunk had four or five people all together in a meeting. And so one of the chunks of the company was getting their data off of machines and another is from, we'll say, processes that happen. And then yet another is from the literature and data sets that exist from from data that's been collected, basically.
So but yeah, are you getting it? I'm inferring that maybe you're thinking that you can't even start this type of work unless you have data together in the first place. Is that what you're thinking?
Oh, yes, because currently we are trying to use call transcripts for the ILP model. And it's a little bit challenging because the quality of these transcripts is not very good. So I'm wondering what type of materials do you start with?
Well, but. Right, so I yeah, I can't get into the specifics of that, but I'm just wondering if you've thought about so you said call transcripts, but I'm just wondering how are you generating the transcripts? I'm hoping it's not human beings writing it.
No, it's generated by a system.
Yeah, so I'm just wondering if you've been able to consider upgrading the tech, the speech to text technology that you're using, perhaps that could help increase the quality because, I mean, it's that that technology is evolving at an alarming rate. Well, it would be alarming if it weren't working for us as a human species. Right, because that's what Google is using. I don't know if you're using Google to extract text from the speech, but there's a there's a variety of technologies that are doing that. Can you comment on that? Or is that also proprietary on your side?
Oh, probably I cannot share details either. But anyway, thank you for your suggestion. And if needed, we can talk later.
Yeah, cool. So we just had a great conversation about the thing that you do and the thing I do that the things. Cool. All the things. Thank you. All the things. Thank you, Rachel. Yeah.
I saw an anonymous question that said, is it OK to post a LinkedIn URL or mention that you're looking for data positions? And absolutely. Feel free to put anything into the chat that you want. And if people are hiring for certain positions, if you want to share those job roles there, feel free to share that, too.
An anonymous question for you, Regis, was. Question from all clinical pharma management, do you have a validated server to run R and can it replace a SAS validated server? I will speak in hypotheticals. Great question. Validated servers are necessary. They should exist and setting them up is complicated. That's the end.
Another question, another question was, I think this is from Bill in the chat who is commenting on the novel approach. Have you published parts of it online anywhere and submitted it to The New Yorker? Of what? On your novel that you're working on. Oh, no, not yet.
Because here's the thing that's that's I love it, but also it slows things down. The world, all these smart people that are on the call and a bunch of other people keep coming up with other. Technologies that make for great narrative fodder, so. I, I keep getting tempted to like, oh, well, that would be a great plot point. Now, this would be a great plot point to like the the there's a YouTube account, I think, where a guy knows how to act like Keanu Reeves. And so they deep fake his face and just think about all of the potential espionage and whatever that can occur. So it's data science driven. Interesting plot points.
Marlene, I see you have your hand raised, you want to jump in? Yeah, I think the answer to your problem, Regis, by the way, hi, and it's been pretty fun listening to your chat. I think the answer is instead of a book, because the lead time is so long to get that published is a comic, right? Something that's quicker to spin out and it's a little more time sensitive with what's coming out. Because my fear is you get done publishing and it's usually a year or two down the line. And almost some of the conversation is obsolete because so many, especially this particular industry, right? Everything rolls out so fast. But a graphic novel or comic would be really rad. So food for thought. That's a really great point.
Community topics and the Shiny polling app
Regis, something I was wondering about with the events you have every other week, what is something that you found that people are really interested in learning more about? Or to be able to suggest topics to? So I have a, I built a Shiny app for polling people so that I could look at the distribution of interest cross-sectional or integrated cross-section with the department distribution across our company. And yeah, the NLP and machine learning are at the top of people's interest.
So that is without going into specific detail, because if I comment on what people are super, super interested in, I don't want to make any, I'm not making any public statement about any, what my company is working on. So, but I mean, machine learning is what everyone is just salivating over.
Achieving scalability through standardization
So how do you go about like finding people to actually speak on those topics? Like, are you, you're actively reaching out to people that you think? Yeah, so one of the things I do, and I hope this doesn't make anyone who's in the group who may be in attendance here, not do this, but so I've got what I, what I do is I maintain everyone is special in case anyone watches this video, you're all, everyone's special, but in order to achieve scale of business, to achieve scalability, I do have to standardize things a bit. So I have on, cause we use confluence. So on the wiki that I've made, I have a lockdown page where I have like some standardized messages that I send out to people at different phases of things. And I do customize and I read what people say about themselves and I'm like, okay, this could be useful here. That could be useful there.
So one of the things that I do is I kind of have this whole ingestion pipeline that I maintain. So to identify new members of the community and then of those members who can help the community learn more about what the blockers are to the current responsibilities they have for their job. So, I mean, all companies have a directory where you can learn about who exists at the company. You can see people's job titles and stuff. So one of the things that I do is I can look into the directory and I can look for like data or data science or whatever. And then the people that come out in the results, like when I search for those terms, I look to see if they're already in my Microsoft Teams group. And because I'm thinking that they could be potential new people who could benefit from being in the community. They might be able to do their job better and also help other people do their job better if they can connect to others who have similar needs or can offer similar things.
So I look in the company directory and then, and I do it on a semi-regular basis. So I can really help to onboard people who have just joined the company or maybe moved into a data science type role. So then I reach out to these people and I use my standardized thing like, hey, I just wanted you to know this group exists. And I only send one email. I don't spam people. And I say like, hey, this community was built just for people like you. This is what we do. Here's a link to the wiki that has the links of all the recorded, because we have a growing library of all the conversations, just like the data science hangouts that you have here. So I send that link out to them and I say, hey, this is something that may be of use to you. Let me know if you're interested in, I can add you and look at all the conversations you've already had. Then maybe you may be able to already do something with it yourself already.
So then of the subset, because it's not a hundred percent, sometimes it's like 20% that respond and they say, yes, this is amazing. Thank you for letting me know. I just came to the company. I didn't even know if there was a community and I'm also doing it from my own perspective of when I joined, it was hard to see who else may be able to, I'd be able to speak the same language about and not the language where, when you talk to someone, their eyes glaze over, it's not for the non computational, it's for the computational people. So when they respond and then they say, yes, they're interested, then I add them to the group, but I have another thing in my wiki for myself, my lockdown page, that is the intro, the welcome words. So I add them to the teams group, then I add them to the active directory group, the listserv, and then I also forward the invitation to the ongoing gatherings.
So every other Friday and in the invitation text, I paste in, welcome to the group. Please introduce yourself in the introduction section. And I include a link, which I use a tiny URL for, so, because I'm using the full URL from Confluence and it was chopping things off. But now that I use a tiny URL, when they click it, it takes them directly into not just the team for the group, but the channel of introductions. So since I switched over, I've gotten a lot higher percentage of people who've been added to the group to start introducing themselves and say what they work on.
So then I could see what people are working on. And then I integrate that with my understanding of what the output is, what people have voted on that they need help with from the Shiny app. That's the poll for what people are interested in. And that's an open-ended thing too, by the way, it's not, that's fixed set of things you can in Shiny, you can configure it so that you can actually add things in. And that's got a MySQL backend. So whatever people pick from the dropdown menu, it stores everything in the backend. And then since I'm using a global object, I'm all over the place, I apologize. But since I'm using a global object, whenever anyone updates something, it automatically updates the global object and that propagates it down. So it updates everybody's plot, regardless of what computer you're looking at.
But I use that and then I integrate it with what people say in their introduction. And then I reach out to them and say, Hey, would you be interested in being one of the guests for an upcoming thing? And then I send them the link and they can see there's no pressure because I'm a year out. I've got someone every other week. So it's like, Hey, would this be something? And then I kind of just am gently having a conversation over time. And then sometimes people drop out and then I reach out to those people and say, Hey, would you be able to switch?
So I think I have all of this in my head all the time, but I feel like if I, if I, this is my normal state, I think less than this, I would start. I think I'm probably like one of those, those puppies who you probably shouldn't leave at home with nothing to do because I'll just be like, so yeah, this is great. That's how I do it.
The roots of community building
I love to see how excited you get about this too. And I, I mean, I want to like plus a million to what Jared said in the chat, I'm trying to scroll up, but Jared said, Oh my God, you need to teach a class for organizations for onboarding. But I'm curious, where have, where did you learn all of this about community building and how can we all learn this from, from that?
I don't think I'm going to have a satisfying answer. I don't want to say the answer in my head. Cause it's super not satisfying. I, okay. So it just comes naturally. It's not fair, but it comes naturally.
My, one of my favorite places to be is an airport, not because of the lines, but because like for me, because, okay, let me back up what I try to do for my, and it does feed into all of this stuff because I'm trying to enable the things that I'm trying to make possible through my group. So the purpose of the group that I've created is to help people build their own careers and also help the company help patients with data science more effectively. But you can learn from other people about how you can build your own career. If you see how they built their careers, sometimes you may be able to plug and play different things, mix and match. And also hearing how other people have done, the, the data science implementation in their own respective projects, you may be able to also mix and match as well. Kind of like this series.
So the purpose of the group that I've created is to help people build their own careers and also help the company help patients with data science more effectively.
But naturally the way I see the world though, is like a combination of the matrix and Tron in that in the matrix, you see, like all the, you know, the, the, the guys that were looking at the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the guys that were looking at the, the coding in the different, the ship. And they could see how everything was like connected and running and causality and all of that stuff. And in Tron, the way the light cycles go around. And then you kind of like see the story path. I mean, I'm thinking in they're called world lines from my study and from the novel. It's an objects, four dimensional path in space time.
But that's how I see everyone, though, every single person that I just walk around in the world, it's like, they have a Tron light, cycle path behind them. And so I want to understand about their light cycle path, their world line. And then, because I just think it's interesting. Everyone has a different path. I just think it's interesting to know but also it can help other people. So then I try to think about how people, because I'm fusing that with the data driven decision-making that I tried to make it possible for other people to learn about. And in order for people to make decisions, they need the right information that can help increase their confidence beyond the threshold of stuff.
So I am just, it's, my brain is kind of always running in the background. How do I reduce the activation energy so that people can make decisions to get engaged with the, the organization and also what frustrated me in the past? How did, why did I get frustrated when I was at a meeting and I was like, I don't, I don't care about this. This is not relevant to my life. So we have over 300 people in the group, but we don't always have 300 people show up. We have like 50 to 60, sometimes a little bit more. And that's because part of the process that I do, because I'm always thinking about data-driven decision-making, but from a mushy side.
Because actually my father, both of my parents are guidance counselors, but my father's more of the emotional lifestyle kind for students. And my mother's more like the hardcore organizational career stuff. So that's the example I saw my whole life. So because of that, I'm just naturally doing both things. Um, so every single thing that I do is geared towards emotionally and logistically D decreasing the burden that people face when they're trying to decide, should I do this or not? And so my brain, it just things pop into my head. I just absorb things from TV shows, books, whatever I am. I hear other people and I'm like, Oh, I could plug this in. I could plug that in. My brain is like a permanent graph database where, um, I just propagate everything at all times. My wife isn't the biggest fan of not having separation of concerns in my brain, but
thank you for that visualization too. And, and honestly, it's actually kind of validating for me to hear that other people's brains work like that too. Um, but Mike, I see you had asked a question, uh, in the chat a bit earlier. You want to jump in? Oh, um, thanks, we just, I, I'm loving what you're saying here is it's amazing stuff. Um, when you were talking, I, and I apologies, I, um, jumped in a bit late. I may have missed this, but when you're collating all of that organizational information, are you building a knowledge graph?
What do you mean by collating organizational information? Well, you were talking about, um, the, in the onboarding process, being able to identify all the different people, uh, within an organization, right? That would be a good idea to store that information in knowledge graphs. Um, I am identifying candidates, uh, for inclusion on the basis of what their job title is, which is already something that exists. But I think it would be a good idea to get a bit more information on what their skill sets actually are. And of course that changes over time as well, but at least from a given timestamp, what is their set of skills? That's, um, that's a need that I've, I'm pretty sure exists at a lot of different institutions. And I've heard it two different points, um, at least in the company that it would be great if there were some sort of, uh, searchable repository of skill sets that exist. Because we don't always need to like go outside where we could have someone who's already an expert, but how do you discover that person? Is that what you are kind of getting at?
So person X has skill Y or uses tool X and, you know, speaks to person, person Zed, you know, so now who uses tool A. So now you've got these connections between individuals as opposed to just kind of dry data of person X knows thing Y.
Yeah, no, that makes complete sense. But now doing an emotion check that feels a little weird without consent though. I don't think I would do it without consent, but also, um, I think the, the schema that you were just talking about, because now we're kind of getting into the nitty gritty of the different types of graph databases, the triple store versus an RDF. And I would probably have the person attributes on the person node, but then have an edge connecting to a tool that they use. Um, but yeah, that type of thing would, yeah, if there's consent, I think that would be great if, if we had some sort of registry that would allow us to do that.
Yeah, I think that would be great if we had some sort of registry that voluntarily, but of course, now I'm also doing the, what's in it for me with LinkedIn. Well, yeah, we do have that in LinkedIn, but I don't know how open the API is. I don't know what sorts of access there would be, but also I'm just thinking the what's in it for me scenario, the walk away, not Regis, but what's in it for the people who would voluntarily provide their information that would have to be substantiated. I don't know how efficient of a threshold to establish an edge connecting to nodes, but I think there's definitely something there.
So that, you know, the kind of thing I'm thinking is, you know, often when you reach out and you say, okay, who knows R, right? That's one question, but then it's almost the people on off the fringes of that, who are kind of like these people know people who know R. And so you can then say, let's connect people together who either are in the same department or, you know, can find other folks kind of on that loose connection sense.
Yeah, no, no, it makes perfect sense. No, no, no. It makes perfect sense. I'm actually involved in a big project that I can't comment on right now, but I'm super excited about it. And the only reason why I'm involved in this project is because somebody who's not a data scientist knew me and the stuff that I worked on and they saw someone else struggling. And they were like, maybe Regis can help.
And then, and maybe I'm biased, right? Because I'm, I'm thinking everyone comes through me, you know, Mike's the R guy. So everything funnels in, whereas if we said, look, you know, there's, there's 50, a hundred people out there who knew R well enough, better than me that, you know, we can link you guys up with someone in your neighborhood who knows that skill.
Well, see, I think that's one use that you can have of a team or whatever type of Slack or some sort of organization. One thing I have is in the Microsoft team for the group that I have is a channel for debugging and collaboration. The collaboration doesn't really get used as much, but the debugging one does. And then also the group that I was in, in the R&D side of the company, we started using JIRA to receive requests because we basically became not just full stack data scientists, but full service data scientists. So going all the way from the DevOps end of the spectrum to explanation, presentations of, of really complex topics to non-computational people. And not just Shiny apps or R Markdown documents or, or Python notebooks, but actual presentations. So, and everything in between.
So yeah, it sounds like you're going on a similar path to what I did before. And it's not scalable. It's not sustainable. But if there is already momentum, I don't honestly fully know how to control that. If anyone else knows, please say it in the chat. I don't know how, because if everyone thinks you're the wizard and that's the challenge, that's, this is one of the reasons why I have the community together because non, I'm not going to say muggles, even though my wife is a potter head and this looks like a Harry Potter. But the, the uninitiated sometimes think that all magics are equal and that's for the non-computational people. And then for the computational people, if they, if everyone else is like, oh, you're the person who's standing out, therefore I will ask you for everything. Both happened to me and the small group of people that I was on in the early years.
Well, I guess one thing is if you have a system like JIRA and I see Jeff likes JIRA too. If you have a system like JIRA, you can kind of force an activation energy to your own benefit because if other people have to jump through a hoop, they may say, ah, the hoop isn't worth it. If it's really burning, then those that get through give you an opportunity to contribute more to the institution that you're a part of.
Community building at scale
Can I ask another, have Eric ask another question that's kind of related to this topic too. Eric, do you want to jump in on the pipeline question? Absolutely. First, Tron reference and the data science hangout, my wife's complete, so nailed it, Regis. So as someone who can fully appreciate making, you might say sometimes over-engineered pipelines for certain operations, I really liked what you've been talking about here, but first, did this, like, are you responsible for all this solo and are you getting help to build all this, either from like IT groups or others that are enthusiastic about community building for what you're doing at Regeneron? And I guess my related question is, is this part of your recognized kind of role at the company or at Regeneron? And is that being recognized, do you think, to your liking?
Um, I, that's a lot of different questions. Um, so I was, I was getting help on the other side of the company. Um, when I was, cause there's, there's multiple pieces also of the effort, right? There's the platform and supporting the platform in terms of DevOps, supporting the platform in terms of the users that are data scientists, supporting the platform in terms of users that are non-data scientists. Um, and then there is also the, the recognition from above, um, different parts of different companies may recognize things differently.
I am on the side of the company that understands the challenges associated with it. Um, and I got a thumbs up from the leadership when I posted this thing on my LinkedIn. Um, so that's great. Um, I got a literal clicked thumbs up. Um, I'm not any more responsible for implementing at a DevOps level. And supporting fellow data scientists in my current role. So it's much more scalable. And basically I have a cheat code. I'm not a gamer, but I know that's a phrase, um, in the wiki. Um, it, because basically, I mean, you know, the rule, right? Like it's a, it's a soft rule, but it's kind of true. When you're programming, if you do something twice, you might want to make a function for it. If you do it 10 times, you definitely make a function for it. Cause now you're at fool me 10 times if you don't do it. So basically I just made a function in Confluence, not a literal function, but it is a life function. Um, and I'm able to copy and paste and it is a relatively scalable because, um, I try, I customize things so that it can be done repeatedly. Um, but we had a team that was supporting it and I was able to hire people under me when I was on the other side of the company, when I had a broader set of responsibilities. Um, but now I don't have a team right now, um, under me.
Um, but we're getting things set up again to be kind of similar to the other side of the company, but the responsibilities are a bit different. So that's in terms of the environment. The ecosystem is relatively scalable. Um, and the leadership is aware of it and likes the fact that I'm able to, um, connect the dots with my colleagues around the company. Um, because it helps others, but also it helps us. Because if we can learn about efforts that are being done elsewhere, we can avoid reinventing the wheel and we can save time and we can achieve a more effective cost avoidance. So there is backing from the leadership in this iteration of the responsibilities that I maintain.
Showing the value of community
Okay, I have a question to add on to that for people who are starting to be this person within their organization. How can they start to show more of the value? Of community and, like, elevate. More of those, like, examples of connecting people across an organization.
So, I mean, since I was, since my memories begin, so, like, I think 6 years old or something, it's always been the case. And I'm glad that so many people are still on the call and paying attention. I'm always working on editing and, and, uh, but this is also our community. So it's easier to understand. But since as long as I can remember, I've been familiar with people's eyes glazing over when I explain things in adults, people, like, 10 times my age when I was a little kid, all the way up until now. That is now I don't think it's going to go away. I mean, maybe if chat becomes even better, and then we've become obsolete. Okay, maybe, but there's so many different ways. I don't think that's going to go away.
I think the way to prove the, um, the contribution, the benefit of a community is I just, I use the word stuck and the other word unstuck because that's really the only thing the business, any business, any organization that's not even a business cares about. There's a bunch of stuff that the organization has to do. There are obstacles in the way they are stuck. How do they get unstuck? That's all data science is for really. Well, you can probably do it for exploration and a bunch of other stuff, but you can argue that maybe that's also for going from stuck to unstuck. But the non computational people who have the power to bestow upon you, uh, promotions and more money or whatever you value more, more influence over the company to make other decisions. They really care about what is the value of the institution. If it's a business, business value going from stuck to unstuck.
I think the way to prove the, um, the contribution, the benefit of a community is I just, I use the word stuck and the other word unstuck because that's really the only thing the business, any business, any organization that's not even a business cares about.
Um, to, I forgot who was asking the thing earlier about a graph database. It doesn't necessarily need to be instantiated in a graph database, but if you can show how the existence of that community propagates through the network that exists in the institution. Um, to such an extent that something goes from stuck to unstuck and you can mark that you can measure it. Then you can say this only happened because of me. I mean, there are other types of organizations, which also should exist too, but sometimes they're based on like interest. But interest is a lot more difficult to prove the benefit of then for lack of a better word, pain focused groups.
Like if your community that's gathered around pulling thorns out of your colleagues arms, then you can be like, we pulled 15 thorns out or we taught another group how to pill, how to pull a hundred thorns out because we were forced multipliers. That I think is the way to prove the impact of the community. I'm not 100% sure it's going to work because there's still going to be people with eyes getting glazed over. But if you say, what do you hate about your job, you person with power that I need to prove the use of this community of too. And then if you understand that, or maybe don't even ask them, but you may know because they're probably high enough profile people at your institution that you know what their values are likely to be. If you can think about how the efforts that you're doing propagate to the reduction of pain points for those people and enable them to go from stuck to unstuck, then collect those things and do like what I did with the showcasing. You could show like, hey, people in this organization did this, this, this, this, and this. Then I think that could be a way to prove impact.
Thank you for that. Pankti, I realized we were asking the same question, I think, at the same time. So I wanted to make sure that you didn't have any follow up that you wanted to add in there too. I know it's a great question. And I think, first of all, I'm really enjoying this discussion because I'm doing something or starting something very similar in my organization.
And one of the things, you know, I am kind of concerned and upcoming like I although I'm really excited about it. I didn't know we talked about a lot of value of having community, but ultimately, you know, I think, please, you know, the company I work for it is extremely you know, result driven and, you know, for the right means. And, you know, people have communicating this complexity to potentially management that's not really, you know, as involved and, you know, understands the data science, or the value as much. And so I think it feels like more proving at a time, you know, at the starting point like why this adds value. And yeah, there is certainly a lot of you know opportunities that you know we can, the people can talk about in a showcase but ultimately, that also adds a lot more new opportunities that have been found, which is great, but then it comes to like, who takes on those responsibly do we have you know enough bandwidth to take those on. And at the same time like how do we kind of showcase like this community is actually working, it's worth people's you know our there's a month or a bi weekly to show up to, and, like, I feel like sometimes it could be difficult to communicate that in more guests to a management.
But, you know, but I think I'm just wondering like there's specific metrics that people use or words you know that that kind of really, we can use to evaluate whether success or not, because after let's say six to eight months. Like, yeah, maybe same people are showing up. But, like, I don't know if you had any additional like I know you talked a lot about it. So, so like I. Maybe the world's different and maybe you're going to feel differently from how I felt. But it starts off being awesome because as a people person only in context where people are being nice I'm not a people person when people are not nice.
It really stinks. So, I think, from the get go, you should be thinking about how to indicate to the leadership. Actually, funky Can you comment on what you're hoping though, because I, I want to kind of respond to what you're saying, but the same way that.
And in my opinion, and also I'm in a pharmaceutical company, which operates because we know that people have things in common we have standard hardware. And so, if you have standard hardware you can fix things but also standard hardware can break for the same reason, but I accepted a long time ago that I am kind of like a biological robot.
So you need to think about data driven decision making, if you want the leadership to make decision x in response to your action why, what is the, the causal relationship that will achieve that. What are you hoping will happen.
I'm honestly hoping at this point that we actually go to a place where we are more driven by data, especially you know in a scientific like, you know, we obviously know people look at data to make decisions, but not to the extent that you know, it could be like that's like a simple when right there. But you can people to go ahead.
And then the second part is, you know, there is, we have so much data available, but we almost feel like we're under utilizing it, you know, that's why we have started this team, where we can start using it, but you know in the team, you know we have folks from various backgrounds, but, you know, even getting like there is ideas but there's like there's too much at the same that there's not enough. So it's like how do you even you know take what is important, you know, showcase your wins right away to show that hey, this is important, we should start like showed leadership like or like an imagining like this is important we need like more resources because ultimately that's where it's going to come down to.
And what we're doing, like you know how we're leveraging data or how we're under utilizing it kind of showcase that and I'm hoping that through this sort of community, we engage on more of the discussion because you know, I know what I know, or the team knows what we know we know we don't know what the opportunities are that we don't know every aspect of the business.
To me, to our team, that would be kind of the win. But I guess you know I think it's, I guess it's tough to showcase that, like it feels like it's tough to showcase that.
So I still feel though that there's like two kinds of scenarios that you're talking about. One, are you preaching to the choir, or are you preaching to the preacher?
Because, like, it sounds like you've got three KPIs that are measurable, like key performance indicators. One is you want to use data more. One is you want to use more data. And then the other is you want to use data more effectively. So those are potentially measurable outcomes that you can determine if you have achieved from the outset.
And then you can work backwards to see how you would get to that point and it feels like there's probably an emotional connection. If people are content to just say, well, I already know XYZ domain so I don't need to use any other data.
That is an important hump to get over. And perhaps demonstrating to some of the decision makers that are above the groups that, I mean, unless it's not this case, I don't know what company you're at, but if they're above the people who are the choir already, that are already familiar with the advantages of data driven decision making and focusing truly on the data rather than opinions and anecdotal evidence. Figuring out how to achieve those three KPIs that you said, and the simplest one might be just use more data.
No, sorry, use data more. Use more data comes with it the challenges of trying to identify how to extract that data and get it in the right format. And you've maybe seen in a lot of articles that talk about a big part of the data scientist job, 80% of that is getting data ready and then 20% is actually doing the data science.
But if you want to start somewhere, I think it may make sense to try to use data more rather than use more data and then use data more effectively. That sounds like an optimization aspect. But if you just say, use data more as your initial step, then you could say, okay, this is, you could try to identify pain points.
I'm just trying to think through here, almost like consulting here, but because I think the other two are higher hanging fruit. The lowest hanging maybe is use data more and of use data more, which is a project that if you use the data that you already have, you might be able to say more incontrovertibly, this is the decision that you should make, which option would be best for going from stuck to unstuck most easily in that easiest to obtain KPI.
I'm just trying to think through here, almost like consulting here, but because I think the other two are higher hanging fruit. The lowest hanging maybe is use data more and of use data more, which is a project that if you use the data that you already have, you might be able to say more incontrovertibly, this is the decision that you should make, which option would be best for going from stuck to unstuck most easily in that easiest to obtain KPI.
No, that's a really good analogy and really good way to navigate this.
Now, thanks for your insight. I really like definitely gonna, you know, use this approach.
Closing questions and onboarding
We were talking about just like starting from scratch. So if anybody wants to stay on and just chat with us, we'll, we'll be there. But we just one other question, Libby. Sorry, I realized I missed your question earlier. Do you want to jump in Libby?
Sure, I have no idea what I asked.
It was about if you could improve or solve.
Oh, yeah. Yeah. If you could improve or solve one thing about your industry, what would it be? Because mine is onboarding, which you then immediately talked about. And I was like, huh, maybe his is onboarding too.
I would love to be able to give you a quick answer, but there's only two minutes, and I don't fully know, because there's, there's like five chunks of the company that I'm in. And I'm getting a stronger and stronger hunch that each chunk is basically a different company, because this side of it, I haven't been here nine months yet, but it's, it's like a completely different company from the other side.
But it actually may fit into the the three KPIs that I was talking about. I think one is like, be more comfortable with data driven decision making. That's a challenge.
Currently, because there's a lot of, I don't know if you could tell from the slight tinge in my voice, the, the feeling of like, well, I know what I mean. So that's enough. And it would be great if I knew how to improve on how to communicate with other people.
That's enough. And it would be great if I knew how to improve on how to communicate to people, such that they're emotionally okay with saying, oh, maybe I should look at the data instead of trusting my gut. I mean, people should trust their, trusting one's gut is important, but your gut is informed by like common sense.
And common sense is common to a subset of whatever your world line is that led you to that Tron point. So, but you have to factor in the entire 3D chess area of Tron, you know, like you have to factor in everybody's or as many perspectives as possible.
And so data-driven decision-making enables the abstraction to a broader set of perspectives. And that's hard. If, if I could figure out how to do that, get over it. I see it's, I see it's an emotional hump for other people. And I wish I could improve on that. And maybe it's also applicable to the other side too. It's, it's people and their comfort with things.
That sounds great. Sounds sounds like a little bit less ego for everybody. Right.
Emotion is more ambiguous, but ego is more accurate.
Regis, are you running to a meeting right now?
Um, I am not. I have a thing in a little bit though. Okay. Can I ask, can I have Keelan ask one last question that I see was in the chat.
Data science and mental health
This has been a really interesting chat but you mentioned earlier that both your parents are guidance counselors and your fascination in humans clearly. And obviously data science I was just wondering if you've ever done any projects on the mental health side or with data science or if you have any interest to.
One thing I don't like is working in areas of ambiguity. And I think mental health has a lot of ambiguity, but also, it really needs a lot of attention.
I mean, talk to anyone who's gotten a PhD and you will see a little bit of darkness. It beats you a lot. I would say nine out of 10 people who've gone through that probably have some mental health things that need to be addressed.
But it would be difficult to get a clean data set off of that I know that's not what you're asking me about but my brain automatically goes to like implementationally how would that work but without a doubt, that's something that needs assistance. And I probably would be interested in that I don't. I'm not commenting on whether or not my institution is involved in that kind of thing but I would be. I don't know, like, I might be interested if if people wanted to collaborate, like offline or something.
So when you're talking about being stuck and unstuck. I'm like relating that to humans and those who get mental illnesses are kind of in that that stuck zone and we can use data science to overcome that but yeah if you're interested, I can send you a message
Yeah, that would be awesome. I mean, do you know if there's sequencing or expression data genome data that's associated associated to that no I wish I wish. Okay, well what's, I mean I don't know what types of applications that are for the reason I asked is because my in grad school.
So I'm just wondering about the potential to integrate and then the methods of assaying it because it needs to be repeatable, because if it's not, then you can't really trust the data if it's if it's measured inconsistently across the set. So, yeah, I'm just wondering what what the applications of that type of thing are.
Yeah, again the data set there's there's lots about like childhood information demographics, stuff like that. So, and like I feel like classically, a lot of psychological problems are actually it's not much genetics that's what I learned in university anyway I'm a psych major.
I think the most like genetically passed on mental illness would be schizophrenia and that's only like 20 something percent. So, yeah, I don't know I've been working with this data set for a while and I have so many ideas but yeah.
This is the thing right like it's the nature nurture combo and the nature I think is maybe easier even though it's extremely complex than the nurture aspect. And if it's, if the cause is different, or varies a lot because you have to like, look at the world line I said it like five times now but it is that's what you're talking about right it's not genetic, then it's because whatever happened to them when they went over the speed bump 15 light squares ago.
So, I don't know how you would. It needs to be normalized but how can you normalize something that is not consistent across the data set, that's a real challenge, or you could make inferences that are at like a really broad level, but then would that be sufficient to suggest actions.
Yeah, I'll find it real quick. Okay, awesome.
I know you, I think you gave me a LinkedIn as the best place to follow up with you is that best. Yeah.
I can put it in the chat to the. Yeah, my LinkedIn is just got it right ahead of you.
Also, I have YouTube lets you get your own at at like the at time. So I was like, I wonder if data driven decision making is available and it is I mean it's a lot of letters, but like youtube.com slash at data driven decision making, I own, which is kind of cool.