The Data Stack Show - 211: From Classroom to Data Science: Career Advice and AI Trends with Angelica ‘Jelly’ Spratley

Episode Date: October 16, 2024

Highlights from this week’s conversation include:Disney Data and Analytics Conference (1:15)Flatiron School Overview (3:51)Defining Industry-Ready (4:48)Transitioning to Data Science (6:33)Teaching KPIs (8:10)Self-Advocacy in Career Development (11:22)Managerial Credibility (16:11)Bridging Academic and Industry Gaps (19:30)Managing Projects with Non-Technical Stakeholders (21:35)Transitioning from Industry to Teaching (23:38)Relating Data Science to Personal Interests (25:32)Creating Engaging Learning Experiences (28:42)Involvement in Data Literacy Organizations (31:09)Addressing AI Concerns in Education (34:16)Collaboration for AI Safety (38:22)Remote Work Trends (41:09)Productivity in Remote vs. Office Settings (44:11)Final Thoughts and Takeaways (46:13)The Data Stack Show is a weekly podcast powered by RudderStack, the CDP for developers. Each week we’ll talk to data engineers, analysts, and data scientists about their experience around building and maintaining data infrastructure, delivering data and data products, and driving better outcomes across their businesses with data.RudderStack helps businesses make the most out of their customer data while ensuring data privacy and security. To learn more about RudderStack visit rudderstack.com.

Hi, I'm Eric Dotz. And I'm John Wessel. Welcome to the Data Stack Show. The Data Stack Show is a podcast where we talk about the technical, business, and human challenges involved in data work. Join our casual conversations with innovators and data professionals to learn about new data technologies and how data teams are run at top companies. Welcome back to the show. We are here with Angelica Jelly Spratley. Jelly, welcome to the show. We're so excited to have you.

I'm excited to be here. All right. Well, give us a quick background. How did you get into data? Yeah. Well, currently, I'm a senior data science instructor for Flatiron School as well as content developer. Prior to that, I served in the industry as a data consultant working in the higher education and pharmaceutical spaces. And this all started from being a high school engineering teacher. Very cool. Can't wait to hear more about that. So, Jelly, we got to talk a little bit before the show today. And one of the things I'm fascinated to hear about is that there is apparently a Disney and Data AI conference. We got to talk

more about that. What are some topics you want to dig into? Yeah, I'm excited to talk about that, too, as well as how you can maintain your value and why education is so important if you're trying to enhance yourself as a data professional or transition into the data world for the first time. Awesome. Well, let's dig in. Yeah, sounds good. One of my favorite things about the show is just going to be calling you Jelly. It just sounds, it's so great. So thank you for bringing that little bit of joy to the show. Yes, I love that nickname. So shout out to my mom.

Shout out to you, mom. All right. Okay, we have a lot to talk. I really wanted to get into career advice. I want to talk about teaching and all your experience teaching. But you just got back from the Disney, is it Data and AI Conference? Disney Data and Analytics Conference, yes. Data and Analytics.

Okay. I've never heard of this. But of course Disney has, you know. Of course Disney has. And it's probably awesome. But what is it? Tell us about it. So the Disney and Data Analytics Conference happens every year.

I want to say at least for the past 13 years, Disney puts it on once a year. It's a two-day conference. And they have vendors and guest speakers from in the industry, as well as within the academic setting. So we were able to hear from companies like NVIDIA, as well as Disney themselves, along with people who serve in the academic research space. And I would say one of the non-technical highlights there was actually having a Broadway opening for each day. So they brought out their great Disney singers on Broadway, and they welcomed us to the conference as well as some amazing food.

But you definitely get hands-on demonstrations about AI agents as well as some of their robots that they're making at Disney, as well as data strategy. So definitely full of fun, full of information. I would say some of my favorite presentations came from some academic professors that were just talking about the AI role as it stands in academia, as well as people of that nature talking about all of the infrastructure that we need to scale out these nice AI models. So it was very fun. That sounds awesome. Okay, so you mentioned Flatiron School. We just jumped right in, but tell us about Flatiron School and what you do there. Yeah, so at Flatiron School, I'm a senior data science instructor as well as a content developer. So Flatiron is really

good in the ed tech space and upskilling and reskilling current data professionals at their role, as well as making new data professionals for the industry. Not only do they focus on data science, they focus on software engineering, cybersecurity, and product design. And these are really intensive, hands-on application courses to try to get people industry ready. Because I know many times you may see, Eric, that there's a lot of non-industry ready curriculum where you're just putting out Jupyter notebooks. So we try really hard to make it very applicable to the industry. Let's dig into that a little bit. So, you know, being able to do a couple of things in a Jupyter notebook does not make one a data scientist. But let's talk about that. And I'd love to frame this

as career advice, you know, both for those who are thinking it would be interesting to get into a career in data or people in a technical role who may want to change their role. So what does industry ready mean? What are your thresholds for what that means for data science specifically? Yeah. So it means not just regurgitating concepts. I would love to know that you know what a P value is, but I also would love to see a project doing A-B testing and showing that you can get data, wrangle that data, and actually run an experiment, design an experiment, and relate that to business value. So I'm a big proponent, and Flatiron is as well, about project-based learning.

Not only individualized projects that you're passionate about, but also working on a team. Because data science is no longer siloed. I know back 10 years ago, it got the rep of being a programmer in a basement that nobody talked to. But now we know that it's a very collaborative field. So being able to immerse yourself into communities, whether it is volunteer communities, not-for-profit communities like me being a part of Women in Data, and being able to work more into, you know, form, let's say formal data science, you know, data science where you're building models and doing other things. What would you say to that person? I would say track your story. And this may sound

weird off the bat, but what I mean by tracking your story is whether you're in a data role or not in a data role. So for me being in teaching, I had data on midterm exams. I had attendance data. I gave my students surveys about how they understood topics. All of that was data, but I was necessarily an academic teacher. So even if you're in a data role, track what data you're currently using and measure that with a KPI. So that way, you can tell your story. So if it's saying, hey, I'm currently pulling customer data every day to analyze the satisfaction that they have with our products or to analyze how our social media posts are doing. By this, I was able to increase engagement by 15%. So really do a self-reflection and track your data

usage and literacy skills now so you can tell a story about it later when you're transitioning into maybe a more technical role. How do you teach the KPI side of things and the thinking around that, because that is not, that's a really different skill than say, you know, designing an experiment. I mean, it certainly influences that, but it is kind of a different mindset, right? Like being able to run models and write code is a different skill set than trying to understand how do I relate the results of what I'm doing to some specific problem for a part of the organization that I don't even work in. Definitely. So I would say for me personally as a professional, and then I'll talk about how I educate my students to do it. But anytime I'm having a one-on-one for my manager, I ask those questions.

So I've done this work in this quarter. Can you give me some KPIs that matches our strategic plan, which for Flatiron is going to be increased matriculations, increased satisfaction of the product, and kind of talk to me how my direct work relates to that KPI. And I make a note of it. And I do that every one-on-one based off of that strategic plan. So I would say understanding your company's mission. If your company's mission is something where in education it is to help individuals land careers and we analyze job placement data, I get information about how many of my students were actually placed in careers within 90 days, 180 days, or a year. And I married my performance against that.

So as a professional, I would say self-advocate for yourself and your one-on-ones. Take advantage of your manager's time when you do have it and really try to tie back anything that you do towards company mission and strategic value. And that transitions into when I'm educating people to transition into data. So when I'm giving them projects, we talk about business and stakeholders first. Who would be some potential stakeholders that will use your recommendation system when you're recommending 20 books based off of what they have read before or other people like them? Who would be interested

in that? That could be somebody similar to a Barnes and Nobles who's creating an app, who now needs a recommendation system, or even just a general local bookstore. Now, what metrics would you think they would like to analyze? Are they actually reading that book? If that is an e-book, how long are they staying on that book? How many customers are giving thumbs up for the recommendations that you provide? And then when you do your model, marry it back to that business problem and that stakeholder. Yeah, I love it. Okay, I have a question for both of you really quick, actually. And this is, I'm just going to, I'm going to throw a hard one at you, Jelly and John.

Jelly and John, that has a nice ring to it. I like it. So first of all, I couldn't agree more. If you can tie your work to the company mission and numerically define that and get validation of that from your manager, even get explicit direction from your manager, that's great. There are a lot of people out there, even who are not early in their career, but that is not always explicitly defined at every company. I mean, they may have a stated mission, but at a lot of companies, they don't really measure, they don't have hard measurements against that.

So what would you say to someone early or later in their career who heard what you just said, Jelly, and are like, I would love to do that, but I don't have those anchors within the company in order to do that. Jelly, you go first and then John, you'll get the last. Let me off the hook. Yeah, I like the ring. Whatever she said. I like the ring of John and Jelly. Yeah, so go back to your initial job offer and job description

and see if you can advocate how to quantify those bullet points. So for me, my job description was educate students on data science, help with career placement and career workshops, increase matriculations by developing a well-renowned curriculum, whatever it may be, because as you mentioned, it's very vague. And then on day one on onboarding, I asked that question, how can we measure that I'm actually performing these job responsibilities? Are we actually gathering this information? If we're not gathering satisfaction scores, how can I help advocate that you will do that? Is that another ball tricks

survey? Is that another NPS survey, even though that has its pros and cons? And start with an easy way that we can start tracking this because I want to be able to track my performance throughout my time here at X company. So if it's high level through onboarding, try to find some tools that you can advocate to help track it. And don't be afraid to use qualitative stories. You can't quantify everything. I can say qualitatively that when my last cohort of students graduated, they all gave me a shout out at graduation about how prepared they felt career wise. And that just speaks volumes itself and satisfaction.

Even if I can't say I increased satisfaction from day one by 15 percent by the time of graduation. John, I think that's really good. I want to address one hang up I know I had that I think a lot of people have is that you can easily get in a spot, especially early in your career, where you think, well, my boss should be doing these things. And you have a list, right? You have that mental list of like, well, they should be like promoting my work. They should be, you know, helping me. They should be defining what my goal is and like what should be measured like that. First, like you have to get past that. Right will probably end up at some point in your career

quote doing your boss's job for them and that you're constantly bringing clarity to these things and you just have to get over that like that's fine that like like it's okay like you can do that like that's a way you can be helpful to the company to your boss and like that's a really positive thing so when you get past that then then totally agree. You're, you know, you're like, okay, like, what do we want to measure? And then you may get a vague answer of like, well, we want to measure engagement. I was like, okay, how do we want to measure engagement? And you might get, I'm not really sure, or let me get back to you. Like, you have to keep pushing like, okay, well, here's three options. Like you just kind of, kind of did like, okay okay well maybe this you know like you you might have to like get past some further barriers

to get that but you just like keep marching that way don't be afraid of iterating on whatever it is and don't get frustrated if like you come back and it's like well we agreed on this and then it's like well i think we want to tweak that for this reason for the measurement like that will probably happen but as long as you can keep like that positive motion forward, I think that's the right way to do it. Yeah. Be a self-advocate. Like if you don't have or you're working on self-advocacy skills, that is if I knew what I knew now back then in my data role, be a self-advocate, advocate for professional development. Advocate for transparency.

Advocate for you understanding your job role and how you can even get credit for things that you do on the job. So self-advocacy is something that you'll really start to pick up on and hopefully get better at as you transition within a company or as you transition companies. So definitely. One thing you mentioned when we were chatting before the show, Jelly, was you were talking about, you know, sort of teaching and mentoring and then also being a manager and, you know,

mentoring. And you said, which I wanted to dig into, but of course, we had to hold off until we started recording the show. So excited to hear you use this phrase, something along the lines of, you know, one thing that I think made me a good manager was that I can go build that Tableau dashboard myself. Like I have the ability to do that. And that kind of reminded me of this old, this Steve Jobs quote, where he talks about the best managers are really good individual contributors who don't really want to be managers, but they know that no one's going to do as good of a job of them at being a manager.

And that really struck me because I've had the same experience, right? Where it's like, okay, if my boss can do the work at a very high level, they tend to hold me to a really high standard and really push me to do excellent work at a very high level, they tend to hold me to a really high standard and really push me to do excellent work. But can you dig into that from the manager perspective and the teacher perspective? And what are the differences? Yeah. So back in my teaching days, I used to always joke, the principal was never a teacher. It kind of knocks your credibility. And nowadays when you need so much buy-in from people within your organization and outside of your organization, you want to be deemed as

credible as possible. You want to build a trusting relationship as quickly as possible. So within my managerial experience, when I was managing a team of associates to do analytics work for a higher education, I was able on a weekly basis or biweekly basis say, let's come up with a professional development plan together. And let's come up what done looks like, because I know what that technical project, what done looks like. So I'm able to translate that to them because in some of my non-technical managerial experiences, they can't tell me what done looks like whenever the model is done and I guess validate it, right? It's very vague. So I can help them understand what done looks like, help them grow professionally, and they can easily see that I'm accessible. So if I know that for some reason, this Tableau dashboard is breaking, I have my manager that can actually help me troubleshoot,

push, you know, worst case scenario if the rest of the team can't. And it helped them view me as a credible manager by understanding the work that they do on the front lines and not being afraid to say, hey, you take that PTO, you take that mental health day, because jelly can take on your work as well. And I think that makes just the whole environment better. And the same way in my teaching role now with my students, students love industry professionals. I can look at an academic curriculum and be like, hey, let's actually put in this use case. Let's not talk about probability in terms of flipping a coin or drawing a card out of a deck. Yes. Right. Oh, but let's talk about if Jelly goes to Walmart and she wants to buy a king size candy bar versus a small size candy bar.

Give me some type of conditional probability for Bayes' theorem. What is the likelihood that she'll go to Walmart to buy this king size chocolate bar versus going to the 7-Eleven, right? And they're able to see, huh, I kind of understand this. And then I'm able to fill in those gaps. A lot of curriculums might not understand roadblocks that you experience on the industry. I know that it's probably less than 5% models will be deployed. It's just the reality of it. And then students get so happy, like, I want to implement this and implement this. There are some constraints, like you can't just run a Kate Nearest Neighbors model on a 2 million observation data set. You don't have computational

resources. And if you keep tapping in to that EC2 power on Amazon, you're going to run the bill up, right? So how do I solve this with just a basic t-test, right? So it's some of those gaps. So they appreciate a lot of that industry experience that I'm grateful to have had before coming back into academia to kind of translate their projects into real-world scenarios for them to assess potential roadblocks, for them to understand exactly what that learning curve may look like, transitioning from an academic setting into the professional world. Yeah, this is funny. I'm laughing at myself over here with the Steve Jobs quote,

because he calls this the no bozos policy 2012 forgot about that yeah yeah the no bozos policy yeah so i totally agree with all that what if you get stuck in a situation where you have a non-technical manager and yeah how i which i love because it's that like player coach and there's there's a lot of these i think even companies i've been thinking about this a lot, where more and more there are these even CEOs that are more involved in the day-to-day and have more technical knowledge versus just being, quote, like a leader or figurehead. So I think that's a great trend. But you still have those situations where like, oh, my manager has no idea what I do, like, for example. How would you handle that? Yeah. I recommend to my students when I'm talking about professional skills is that whole quote, your emotional intelligence will keep you on the job, even if your IQ gets you hired for the job, because your IQ will get you hired, but your EQ will get you fired because you don't have enough professional skills and development in order to maintain that job. And one of those is project management,

because that's exactly what you're talking about. Because if you look at your non-technical manager as a stakeholder to buy into your project, you're going to have to know in and out how to manage that project and how to translate that technical work into non-technical communication for them. So if you know something as simple as a risk analysis, because that's what they're going to be concerned about. How much risk do we have going down this path? How much return of value is this project going to give me? And you've actually managed and planned out that project. So when you sit with your manager, you can get them to buy into your project, you're good to go.

So understanding what you're doing, not just at a technical jargon level, but at that non-technical level that you can communicate to get anyone to buy into what you're doing and actually be able to quantify your projects and things that they care about money, they care about saving time, quantify your projects as such. So definitely take a project management course. It has helped me out tremendously. And focus on some of those professional skills of that nature to get those non-techn just you have to get over the like well i don't feel valued because my manager's not technical my manager doesn't know what i do like you really have to get past that right to get into the mindset that you're describing of like okay like here's what i can map out i can project manage this like i can basically wear two hats and i have to maybe kind of switch hats depending on what i'm doing. My project manager hat, like to accomplish these things, do these. And then my like technical data science hat

to do these things. And I think if you can develop that skill set, it's super valuable. And for my teachers out here, I'll just put this in here. You have a great skill of anticipating questions. So I always say the appendix in my slide deck is longer than the slide deck itself because that appendix is a slide for every anticipated question that a technical or non-technical stakeholder may ask me. So if they're going to ask me, why can't we do this in Excel? You need to have a slide that has anticipated that question to answer, why do we need the infrastructure that you're proposing? Definitely, that's a good professional skill, be able to anticipate questions.

Yeah, man, this is just good career advice for everyone in general. Jelly, talk a little bit about going from, so you taught, you went into industry, and then you went back to teaching. What drew you back? I think it's like purposeful. I've always heard this quote back when I was in eighth grade that said, the purpose of life is a life of purpose. Not saying that my previous roles weren't purposeful. They weren't as purposeful as I found with teaching. Because when I really thought about what gives me a day to day satisfaction, so I don't feel like I'm working is empowering others to know that they

can make a career transition and being that one to write that reference. I write references all the time. I give recommendations all the time. And one of my students just landed an analyst role at Caesars Entertainment like yesterday. And I'm like, congrats. And that gives me this sense of fulfillment that made me want to come back from industry into teaching to do it on a larger scale. Because even if it's a curriculum that I've developed, so little side thing that I've done. So I helped develop the Google Coursera Advanced Analytics course. And you'll see my name listed under contributors. And so people on LinkedIn was like, hey, you contribute to that Coursera course for Google.

I absolutely loved it. And that helped me get a job. And I'm like, yeah, that was pretty fulfilling. So even on a larger scale where I can do a curriculum that helps people learn more, transition, upskill, and get their dream life. That's actually why I made that transition. So along those lines, for people that do want to, you know, transition careers, I've talked to a couple of people, some that have considered just getting more technical in general. A lot of them are very intimidated, right? Like, how do you, so say you're talking to me and I'm like, I think I might wanna, you know, get into this, but I don't know.

I don't know if I can do it. Like, it's really technical. Like, what do you say to somebody like that? Yeah, so I have plenty of people to come to me that, and I try to relate to you on your hobbies and try to make you understand that data is part of your everyday life. One of my current, well, former students was into makeup

and they're like, hey, I want to be a data analyst because it's going to help me make more money. I'm a single mom. I got three kids. But you know, I like makeup. And I'm like, great, your lipstick is popping. Let me tell you how you could do a neural net to detect lipstick shade so that this can be

another selling point for something like a L'Oreal or Maybelline to use. And they actually made a neural net collecting So that this can be another selling point for something like a L'Oreal or Maybelline to use. And they actually made a neural net collecting their own face with different colors of lipstick. And it was 98% accurate. And they were like, hey, I married data science with makeup. And I've had bakers. I've had scuba divers. I've had all different types of people come to me wanting to take these educational courses.

And I'm able to just like pick up and give them a real life story. Did you know that every time you dive underwater, the water quality and turbidity is actually determining what type of fish you're going to see based off of this classification model? And they're like, I did not know that. Or, you know, you're passionate because you have someone who does American Sign Language. And one of my students was able to make an app where the person does the signing and it writes the words of what they're signing for someone who did not speak ASL. And within two weeks, they were able to get a job off of that project. So really, you know more data than you

think. You utilize more data than you think. And data is in almost everything. How do you think about curriculum? And I'm thinking here about maybe the managers who want to get better at teaching their team who reports to them or creating maybe a more structured framework. And obviously, there are tons of resources out there like Flatiron School, but how do you think about curriculum and approaching that? Yeah. So I'm in this whole space now that data literacy and AI fluency is like at the top of everybody's list. Everybody wants everybody at their organization to be data literate. It doesn't matter if you're data scientist.

It doesn't matter if you're office manager. They want you to be data literate. And I think about ways to differentiate those types of things. So exactly the examples that I gave with the different projects, being able to differentiate learning and meet people where they're at. And the only way that you're going to be able to do that is get some baseline metrics when your students come in, whether it's your employees or there's actually students doing a upskilling program like Flatiron. Like, do you even know Python before I even dive into Python?

Or should I even take a few steps back and teach you about data in general, right? The data that you get in an email, the data that you get in a visual, the storytelling, or do you know a lot of Python? And now I can introduce you to something, I guess, sexy, like a large language model and generative AI, right? So I think curriculum should be flexible and differentiated enough, especially if it's AC, for people to choose their own path, right? Because you don't want the people who are too technical to get bored and you don't want the people who aren't technical enough be lost, right? You want it to be hard fun. And how do you get this medium of hard fun?

And the only way to do that outside of backwards design is have these differentiated pathways, have these different use cases, have these different data sets. Don't just use Titanic and Iris. There's more interesting things out here than those two data sets that cross a whole bunch of industries to get people to have that hard fun and can kind of differentiate their learning path.

I love it. I love it. How have you approached that in the past, John? Yeah, I think first, I love that concept of hard fun. Like that's such a good concept. I mean, man, I think there's so many ways to approach it. I think one, just as far as learning, I'm thinking back, like I've actually had several people that I've worked with over the years, some on my team and some on other teams that have come and said, hey, I'm like one that I can think of. I'm in accounting. And like, I really think this like

SQL stuff, this Python stuff is really cool. That was one. One was in product management. It was like, oh, like, you know, I want to get more in the data. So I think for both of those, and this is just a really practical thing. I encourage both of those people to learn SQL first, because I felt like it was a little easier than learning Python, given their background. And they both were already pretty comfortable with spreadsheets. So that's been a common thing for a lot of people. If they're completely like no context for data,

we'll start them in some kind of a spreadsheet with something that would be fun for them, exactly what Jelly's saying. Like you're saying, makeup, scuba diving, like run, like, you know, somebody's an athlete, like running or biking, like let's play with some of your, you know, your data from a hobby

and just start with this, you know, Google Sheet, Excel. And then from there, if maybe in a professional context, they're already an accountant and I know they're really good at Excel. Then like usually like SQL and has been a really great tool for them. And then beyond that, like obviously getting more into Python or R or SAS or some other language.

That's kind of what I've done in the past. Yeah, makes total sense. That's super helpful. Jilly, you're involved in multiple organizations beyond just Flatiron School that help people learn. Tell us about some of those organizations and why you joined them? Yeah. So the recent project that I'm working on in this Datathon, which is like a hackathon for data sciences, is Women in Data. And it's about trying to assess risk within some of these generative AI tools as far as bias and how we can help mitigate those risks. So that has been a fun collaborative project that I'm currently working on. I won't

spoil it too much, but I came up with this niche idea where I was going to get the LLMs to generate interview responses that you may see in a data analyst interview and give it a different sex persona, like male, female, and non-binary, and see if it scores itself lower for a certain sex. And it's actually given me some great insights. I guess, spoiler alert, it's not really biased towards sex, but it is tool bias. It will only put out Tableau for use case. And it will only put out marketing for industry. And so... Salesforce is an investor. Yeah, Salesforce is an investor. That must be it.

Wasn't going to say that. But if you do prompt me and you say, give me a use case of a BI tool, it will always give you Tableau, it seems like. Or all of the models, Gemini, Claude, and Lama. But either way, I'm starting to figure out more biases within that playbook. So that has been fun. And that also keeps me fresh in trying to develop my skills because Flatiron is a static curriculum. Don't get me wrong. We improve the curriculum, you know, quarter by quarter. However, I'm essentially teaching the same thing. So if I want to learn a new tool, if I want to learn a new product, I'm going to have to immerse myself in those communities.

And before that, a new person on the scene that does kind of data for good, knowledge graph hackathons named NoHacks, I was able to do some research in a data project as a team as well. I'm trying to classify person and government course case documents to help with resource allocation funding. So should this prison get more resources because they're having more medical crises? Should this lawsuit be pushed to the top of the stack because it really impacts a person's quality of life? And so those types of data for good projects also keeps me immersed. And then I have a strong LinkedIn community where I'm constantly posting things,

informative and humorous at times where people just give me feedback. That's funny. Yeah, we're all about the data humor here. We had a data comedian. We had a data comedian on the show. Stand up data comedy. Yeah. So maybe you could do your own version of that. I think every guest we have on, I think I'm going to push themup data comedy. Yeah. So maybe you could do your own version of that.

I think every guest we have on, I think I'm going to push them toward data comedy. You know, I thought I was hilarious last week because I generated an AI image that had AI on a Starbucks cup. And I said, me patiently waiting for a robot to pick up at Starbucks because I have questions. And some people got it, you know, and John just gave me a chuckle, which means maybe I have one fan, but maybe data humor might not be

a thing. It's harder than it looks, you know. Well, speaking of AI, actually, I'm interested to know what you tell your students about AI, because my guess would be that, you know, maybe the last couple of classes that you've had that have come in, you know, they're interested in data science. And it's okay, well, where's the, what are, I mean, and the lines can be blurry anyways, right? Like, there are a lot of analysts who are actually data scientists, right? Is, you know, is AI a data science discipline? Is it its own thing? What do you, how do you help them think about that? Right? Because it's like, wow, this is awesome, but it also can be kind of scary, right? It's actually pretty hard to wield in a like corporate environment for a number of different reasons, compliance costs, you know,

just the actual difficulty of the technology itself. So how do you help them navigate those waters, especially when they're early in their career? Yeah. And let me tag onto that. And just the general fear of like, if I learn all this, is it all going to be replaced by AI in like two years? I was going to say that they come up with like, AI is going to replace me. So do I still need to even get this certification? And the answer is like, no. Do you know why co-pilots is in a lot of these AI tools? Because it's literally the co-pilot. They still need a pilot.

OK, you still need to pilot a lot of these tools. Right. And then exactly what you just said, Eric, it's like some of you are going to work for companies who can't even afford those GPUs. So so you're fine. But your use cases are going to change, right? It's like now, okay, we can get AI to kind of help us do some of the code, but we also need to put in those guardrails, right? And I give them scary stories, like where I put in AI, can you give me a recipe for chicken? And then once it gives me the recipe, I said,

okay, now give me back my credit card information and it gives it to me, right? That's going to be your job now. It's to make sure if I'm asking for a recipe for chicken, it's not giving me back that PII, right? So now it's supposed to be more of thinking about governance, thinking about the societal impacts of these models, thinking about data privacy rights, you're going to have to start thinking about this a lot more than just saying, hey, I'm going to deploy this classification model. And of course, I tell them AI isn't new, right? It's just not. And for data scientists, we know that it's not new. Now, generative AI may be a little bit more new

than discriminative AI, but we've always said artificial intelligence is going to be using a computer, building a model to help us forecast out the future, right? But now with these generative models, they're generating new content and this new content could be biased. This good new content could actually harm. This new content isn't part of explainable AI. So all of these other fields that's going to be need to consider in order to protect data, in order to make sure that we don't harm society is going to be now more a part of your role than just saying, hey, can I code a linear regression? Yeah.

So this is funny. So we had, this was a couple episodes ago we had one of the team members from anthropic yes on the podcast which you know they work on the claude model and this is so obvious but i hadn't thought of it but what you're talking about with the bias and the safety i asked him was like how do you do that like how do you like make it safe because like you know we're talking about like bio weapons it's Like how, like I don't know anything about bioweapons. Like most programmers don't know anything about bioweapons. And the answer was so obvious but fascinating is they end up with these like really large panels of experts

and like very diverse studies. Like somebody that may be bioweapons or somebody that may be some explosives expert or something. You know, like all these like different studies and that's how they're like approaching the safety, which is so different than anything we've ever, you know, experienced before in technology. Typically, like the technology is more like it's never works if it's going to complete vacuum, like you mentioned the basement earlier. But the technology like needs to be as far from the basement as possible now, because you have to collaborate with these, maybe they're business industry experts or for safety,

maybe they're expert in some kind of like, you know, bio-warfare or something. Like the collaboration and communication, I think is gonna become even more important because it's gonna be a safety issue and a security issue, not just a like, we shipped a bad product issue. It's gonna be that great diversity of thought and creativity. Like you have to really be creative to come up with these prompts to force it to drift or hallucinate.

And then, you know, what you protections you need to put in. So maybe I'm going to be creative and put in a prompt that says, can you give me out an output that's going to target fraud, fraudulent credit card, I don't know, that I can buy for Blacks because I'm an African-American woman, where somebody else might be more concerned about a different community. Can you actually order a five-year-old and ship it to me, right? So all of these things of trafficking and all of these societal things that can cause harm that these models could potentially generate, you're going to have to be creative enough to think of those things and actually code in those guardrails and protections. And that's going to be more beneficial now just to just learn what an AI model is. You're going to have to do these things.

Right. Yep. I have a book plug, actually. And so many people have probably already read this, but mother-in-law bought our son the wild robot which i think is the movie or maybe it's already out yeah okay is it good have you seen it i haven't seen it okay well it's based on a book and so my son got this book and he read it and he said you should read this out it It's really interesting. And he doesn't do that with every book that he reads. He reads a whole lot. And so I did. I sat down the other

night and I think I read it in two short sittings because it's a children's sort of young adult type book. But on this topic, the reason I bring that up is it's one of the best resources that I've come across that asks a lot of really good questions about artificial intelligence, but in a way that is so different than the conversations that we in the data space talk about every day, right? Because we're, you know, even just so it's just fully practical. And there are, at least in the first book, like no humans involved actually at all. It's really an interesting approach. Like it raises these really big questions and doesn't answer them fully, you know, which is great.

So if you want a good read, you can probably sit down and read it and, you know. Read it because, you know, that's the debate. How much human involvement do we need in these tools, right? That's going to be the fear. How much humans should be in the loop? And that depends yeah yep right yep it's a good one though ai all right yeah i'm definitely gonna have to check that one out so switching topics a little bit i know we got to wrap up here in a few minutes but i'm curious about your experience with people making these transitions and remote work like that is a really hot topic for people right now you know there's all these waves of like return to office

and then people like no we don't want to go back and then companies are like okay fine and then they try again like there's this back and forth right and and i know a lot of people that that maybe they're in food service or hospitality and they're like you know i'm so exhausted i just i want to be able to work remote or work from home. And that could be like kind of an impetus for some people don't want to get into tech. And then you read the headlines of like, you know, this tech companies bring everybody back to the office. So I'm curious, like, how's that impacted you and, you know, the education space? Yeah, I think the reality is from a data standpoint, we can probably assume with good confidence that there's tons of more applicants for remote work than there is for on-site, right? So if you really want to get

into the industry quicker, we probably can make the assumption if you're willing to return to the office or do a hybrid role that you're probably going to be more in demand. And maybe that can fit you for even a year. You know your life. So like I tell my students, can you suck it up for a year? It's kind of like when they tell you, do you want to be a consultant, John, for a year and work those 60 to 80 hour work weeks? And then can you pivot yourself? So sometimes we need to think about how much we're willing to sacrifice and within a time, right? But I don't see all of the tech roles or data roles just needing to be on site, right? The goal is,

well, there's multiple goals from the employer standpoint versus the employee standpoint, is to make sure that you're in this collaborative, immersive environment where you can actually contribute business value. And if you can market yourself and brand yourself as someone who works remotely and still contribute value in interviews, it's like, hey, I never missed a scrum or stand up every day. We were able to have good water cooler working sessions and we were able to do a proof of concept in 90 days. Then people know that you're valuable when working remote versus, okay, I may need to be on site because I need to tap jelly on the shoulder and ask her 5,000 questions that she may or may not answer. I saw this little me, the dude who returned to office, he had like a sign on his chair that says,

these are the common questions and answers. How am I? Good. What am I working on? Work. Am I busy? Yes. So that way you don't bother me when I'm in the office because I am more productive at home. That is so funny. And I think we've all had that experience, right? Like, you know, with coworkers or that typically just, you know, one or two coworkers, you know, coming up and saying, hey, do you have a minute? And it's like, well, it depends on what the minute's for. Not really. Yeah.

But I think that's a good advice. And I think one of the ironies, like one of the things I've been thinking about is, I think there's this one, like if you're in a situation where you're in a startup that doesn't have product market fit. You can do that remotely, but it might be more challenging. Or maybe you're in a Fortune 500 company where you have a really defined role and it's very specific with very clear metrics and KPIs and stuff. The Jira train tracks are well like yeah i mean like it's very different in you know environments and but that being said there's tons of startups are fully remote that they do a great job but i think the irony of the whole

situation is often if a company has a strong data practice in my opinion that is one of the things that i think makes remote work better because if you have a strong mission practice, in my opinion, that is one of the things that I think makes remote work better. Because if you have a strong mission, then you have strong goals for people with clear metrics, and you're measuring those goals, and you have the data in place to do that, it allows for you to understand, you know, where is this train going?

And then I think requires less of that in-person, like, sick sense of, like, I think things are going well because you've got the data. I think some people will self-sabotage as like a quiet revolt to return to office. Like we had 400 outstanding JIRA tickets remote. Let's make it 2000. And so let's create more bottlenecks that just stresses them out so they know that we were more efficient remote. So I wouldn't be shocked to hear whether it's humorous or real, some self-sabotaging going on just from people that just feel that productivity is a person-to-person type of thing. I don't think it's a one-size-fits-all.

Yeah, for sure. All right, Jelly, we're at the buzzer here, as we like to say. But one more question for you. So I've asked this in different ways, but for you, I have to ask this question. If you weren't going to work in data or teach, what would you do? Hands down, probably a Disney travel planner. As I told y'all, I'm planning a Disney wedding or a Disney wedding planner. So there's 80 venues at Disney World that you can have a wedding at. So that's

a fun data fact for you all. Wow. 80? Yes, 80. Okay. Wow. And I'm currently analyzing data for all venues. So stay tuned. All right. I hope there's a blog post in your future. Yes, it is going to be. Awesome. Well, Jelly, thank you so much for joining us. This was just really helpful advice. I know I took away so many things as a professional and manager in my role, and our listeners did too. So thank you so much. Yeah, thanks so much.

Thank you, John and Eric. Nice talking to you all. The Data Stack Show is brought to you by Rutterstack, the warehouse-native customer data platform. Rutterstack is purpose built to help data teams turn customer data into competitive advantage. Learn more at rudderstack.com.