#110 Behind the Scenes of Transamerica’s Data Transformation

Hi, I'm Maggie, VP of Curriculum at datacamp, and I know firsthand how overwhelming it can be to keep up with the rapid changes in data and AI. But staying ahead doesn't have to be hard. That's why our curriculum team has curated the best learning content to help you and your team stay ahead. Covering everything from data and AI literacy for busy executives to cutting edge generative AI for developers, and everything in between. Whether you're upskilling yourself or your team, we make learning practical, engaging and flexible. It just works. Power your data and AI transformation journey on datacamp.com. You're listening to Data Framed, a podcast by Data Camp. In this show, you'll hear all the latest trends and insights in data science. Whether you're just getting started in your data career or or you're a data leader looking to scale data driven decisions in your organization. Join us for in depth discussions with data and analytics leaders at the forefront of the data revolution. Let's dive right in.

Welcome to Data Framed. I'm Richie and today we're talking about data transformation programs. Whenever I speak to datacamp's customers, one of the most common conversations goes like this. Hey, we know we need to get better at working with data, and our C suite has finally figured this out too. So now we're going to do a data transformation program, but it's kind of hard and I'm not sure exactly what we need to do. So at datacamp, we spend a lot of time coaching organizations through the details of who needs what data skills in order to modernize their data usage. And I thought rather than have to tell every organization one at a time, let's just hear the story of someone who's been through that transformation process and telling her war stories. Today is Vanessa Gonzalez, the Senior Director of Data and analytics for Machine Learning and Artificial Intelligence at Transamerica, as well as helping Transamerica through their data transformation program. Vanessa is also a senior Data manager, so I'm expecting some great leadership insights too. Hi Vanessa, thank you for joining me today. I'm very excited to chat about what you've been getting up to at Transamerica. So first of all, maybe you can just give us a bit of context about what does Transamerica do?

Hi Richie, thank you so much for having me. So Transamerica, it's a financial institution. We do retirement, we do employee benefits. When you hear about a company that, for example, when you start working there and they offer you a 401k and they offer you some benefits that you can choose from that's what Transamerica does. And then also Transamerica on the other side of the of the coin can sell some products and annuities benefits directly to customers. So we do a little bit of both. We're really well known for the retirement side of it, but also we're getting into a lot of other products like employee benefits and insurance.

Wonderful. And so your job title is data and analytics for ML and AI. So maybe you can just explain a bit more about what your team does.

So I have a team of data scientists and also I have a business system analyst in my team and I work very closely with engineers, with architects. But really what we do do is that we figure out how can we help our business there. It's very exciting and there's a lot of different topics and there are a lot of different ways how we do it. But what we do is that we, we use machine learning. We use AI to create more value to our business. We help them solve problems and we make sure that by doing that they can do their work better. And also they, we can be better with our customers and get them to have better service as well.

And are there any particular business problems that your team's been working on?

Yeah. So we work in many different things and that's the fun part of our job is that it's never the same. So if you ask me today and you ask me a year from now, the projects are going to be completely different. But to give you some ideas of what we do, we focus on four different areas. So everything that we do is to increase retention for customers or to create growth. So really grow our business or improve customer service. So it could be from the call center or it could be on how we do processes and how we automate certain things like reduce the, the time that you wait on the phone for example, or, or if your call is routed, is routed to the right place. And we also try to decrease costs for our business. So depending who we're working with, we're going to be doing different things. But everything we do is going to have a machine learning model that is going to drive these predictions that help our business and then we integrate them into our systems that we already have. Let's say one example is if, if advisors to know who would be a more probable person to be retained, we help them by giving them a prediction of these and then they can call this person and talk to them and figure out how can we help them with the issues or problems they may be having. So that's the type of thing that we do. And we do a lot of other models as well for prioritizing. So, for example, if we want to know which claims may be fraudulent, we can see, okay, these top 10 are the ones that look like more like fraud. So we can do models for that as well.

That's really fascinating. So you mentioned that you have some data scientists in your team, data architects and engineers. So perhaps you can tell me how all these people work together. How are your team structured?

Yeah, sure. So my direct team were more data scientists and business analysts, but we work very closely with the data engineering team, with a third architecture team, with a BI team. So the way we do it is that, as we always say, that machine learning is a team sport. So you need to collaborate with all these teams to make it work. So you're going to have three pieces for every model that you're going to build or every solution that you're going to build. You're going to have the piece where you bring the data in, and then we need the architects there and the engineers to bring that data into cloud, make it available for us to access it. Then we have the data scientists in my team, they're going to be developing those models, they're going to bring the data, manipulate the data, they're going to be working with it, training the models, developing them. Once they're ready for deployment, then we need to work with a DevOps team to make sure of how we're going to deploy the solution. We need to bring that model from development all the way from all the environments, all the way to production. And then there's another piece to it. We need to integrate the results of these models or the output of these models into solutions or applications. So could be Salesforce, it could be just the table in Redshift and the cloud, it could be other solutions like Call Miner that we use also for the call center. So depending where we want the output to be viewed, then we're going to have to work with them and that we need engineers again, DevOps and the architecture team to help us out. So that's how we interact. So we may not have everybody in the same team, but we have to work with all these teams to make it happen. And of course the business, that's the most important piece of or the most important team, because we're really trying to have them explain what they're dealing with, what problems they're having, and they help us through the process also to get feedback of what the results that we're Giving them, and then we tune our models and then we're able to do a little bit more there.

So it really is like a lot of different teams involved just to get answers to these data problems. It's not just data science working in isolation. I like that.

Exactly. So when you think about a data scientist, if you think that they're going to be working, just hiding in a room doing their thing. Well, not really. They need to have a lot of communication with other teams. They need to have a lot of collaboration. So for a day, a good data scientist is going to be somebody that, that loves to collaborate, that loves to work in a team environment. If not, they're not going to be able to develop the same quality models that you could if you integrate with all these teams.

I think that's really useful advice there, that you do need those communication skills. Actually, maybe just continuing on that theme, are there any particular skills like communication or are these other kind of softer skills that you think are important for data scientists?

Yeah, definitely. So one skill that it's not easy to find and is very, very important is not just knowing how to communicate, but also knowing how to translate the very technical to the more everyday type of work. Because you're going to have to be working with business people that have never seen a model or they don't know how it works. So you need to be able to have that communication going back and forth and understanding what they want to tell you, but also being able to share what you're finding and what you want to tell them in the same language. So that translation, it seems that it's easy, but it's not that easy. Sometimes you have to explain a model that is very, very complicated in a very easy way. And sometimes the business has to explain their processes. That may be for them, very obvious to data scientists that they have never been exposed to them. So it's not as obvious as one would think. So that communication skill, definitely important.

Do you have any success stories where that's been done really well in your organization or any maybe some disaster stories where it hasn't worked so well?

No, definitely. So for data scientists, you know how we always say, oh, well, this is the recall and this is the precision of our models? Well, that doesn't go very far with the business because they don't know what recall is or what precision is, or even when we're talking about accuracy or nev score, what are we talking about? So I have a data scientist in my team that he's awesome on that communication with them. So he's able to say, instead of using really the data science terms, he's able to tell to the business. And in this case, it was a model that had to do with natural language processing. And we were talking about how many, like how the model was identifying topics in a call, in a transcription. So he was able to really explain to the business on how accurate was the model by using some easier terms, like saying, okay, of every 100 calls, the model will be able to tell. Tell us 20 times correctly what the topic is. And then it would not be so sure in another 20, 20 times, but in five times there was a goal. So they were. He was really able to explain what we were trying to say with the results of the model or the, like the metrics of the model in a, in a way for the business to comprehend and say, oh, 82% of the times guessing the topic is good for us, it's even better than what we get from our own people doing it. So we're very happy with that number. And then we were. When that conversation went from there, that would be a time that really worked out well. We have tried in another times when we go and just give the metrics and we just get a lot of silence in the room. That's when you know that you have to explain in a different way for everybody in the room to understand what we're trying to say. And, you know, they're not supposed to know machine learning. So we have to be able to say, what value are we going to add by doing it in the terms and in the way that they're more used to? So that's always an interesting conversation, but you learn it and you get good at it. And by practicing and paying attention, you can really get that translation really to a good place.

That just seems so important. I think, like, one of the points you're making there is just if the business people don't understand what you're talking about, then you just have no impact across the rest of the organization.

Exactly.

All right, wonderful. So you've been part of a big data transformation program at Transamerica. Perhaps you can tell me a little bit about what the goals of this data transformation program are.

Yeah, definitely. So when we talked about data, we need to, as time goes by, we need just like a lot of data, but we need to access it in an easier way. We need quick access to it. We need to be able to find the data in one place, and we need to make sure or know that that data that we're going to be using for whatever we're going to be using it for, that it's accurate, that it's complete, that it's timely. So Transamerica, we. We're a company that has been around for many, many years. I think more than 100. And it has been also. It has been formed from acquisitions and it has grown in many ways. It has been restructured many times. So we have many sources of data and we need to make sure that we can access all the data that we have. Also think about a company like ours that we do retire. If you have somebody that starts their 401k when they're in their 30s, they may not start using it until 30 years later. So you have customers that have been with us for 30 years or for 35 years, and that means that we have to keep all the data of all the transactions they have done in their plans through that time or in their. How. What if they have been maybe married and then divorced and then they had kids and. And all the beneficiaries for them have changed through time? So there's a lot of data. What we are doing with the data transformation is really moving all our data from on premise servers to on cloud. And we are trying to modernize ourselves to make sure that we have all the data in one place, that all that data is curated, that it's accessible, that it's really well monitored for security as well. We want to keep our customers protected. We don't want their data just floating everywhere. So we have to make sure that we do all these things. So by doing a data transformation and a digital transformation, it allows us to be a lot better, more careful, and use our data in a better way. As we move our data into cloud, we also make sure that the quality of it is there, that we're looking at how we're using it, that if we have somebody's record in seven places, that we know that those seven records of that person are the same person. So we're doing mastering and identity resolution there. And most of all, we're trying to have the data available and insecure for our customers. So that's just some of the examples of why we're doing the data transformation. But as you can imagine, it's a huge project and it's a very exciting one, for sure.

Absolutely. I mean, I think about the data we have at Data Camp, and the company's been around for, well, almost 10 years at this point, and we already have data from so many different sources in so many different places. So what you're talking about where someone's got a life insurance policy or retirement policy and you've got to manage the data integrity for 30 years before they even start using it, then that does seem like a huge challenge. So can you talk to me a bit about, like, where have you got started with this program? At the beginning, you have data in all different places and you're trying to curate the data. So what was the first step with this?

So the first step, and that was started even before I started in Transamerica, that we started thinking about, okay, what do we need to do to be a more modern organization to keep our data safe, to put it in one place and in the right place? So the first thing is making the decision of, this is what we want, this is important to us. This is going to be part of our strategy. Then from there, then we start to thinking of, okay, how are we going to do this? Because it's huge, it's a huge project. It's not something that you can get done in a day and it's not something that we can say, okay, everybody stop. Everything they're doing. We're going to wait for one year or two years while we do it, and then we continue business. We have to keep the business going. You have to keep both things happening at the same time. And that's also tricky. The second piece, first you start the strategy, you start thinking about how you're going to do it. And then the first step to do it was really creating that architecture, that foundation that, like the little boxes where you're going to put the stuff, right? So you have to figure out what is going to be your architecture in cloud. How are you going to do it? Are you going to bring applications? Are you going to bring just the data? Are you going to bring both? In our case, we are doing both. We're bringing like, the idea is to have, in a, maybe in a year or so, we're going to have everything in cloud, maybe in between one and two years. So we have already brought a lot of applications into cloud. Now we're bringing data. We have about 25%, I would say, of our data is already in cloud. We're going to bring a lot of data this year. We have so much that you have to start thinking as you bring in, okay, what am I going to clean up? What am I going to bring the data from one server and then just shut the server off? But then how many processes are affected in that by moving that data? So just think about reporting if you move the data from point A to point B. Every report that was using data from point A have to be refactored to point B. So it's a lot of pieces happening at the same time and you have to prioritize then what comes first, what comes later, and the sequence of how you're bringing in the data and the applications and everything else. So the first step is really getting that architecture ready, getting that place to start moving things in, making sure that you have the security that you need. How are you going to give access to that data to the applications? Like you really start thinking about that architecture. So our architecture team did an amazing job about thinking about it, getting a lot of knowledge on it, making sure that the way that they're setting the architecture is going to work for our company, because every company is very different. So we cannot just say, oh, maybe Sony did it this way, we should do it the same way. We have to come up with an architecture that works for us and that is going to work for our customers and for the agents that we work with and the companies that we work with. So there's a lot of different moving pieces. Once that is set, then you start bringing things in and you start thinking about, okay, how do I bring them in, for how long I keep them both, or in which cases I just move them, how do I test it, how do I give access to these new pieces? And then once we have all that, then you start to have to think about how do I, I turn off the old and the legacy stuff and just keep the new one. So that's more or less how we're planning and how we're going about it and how we're doing it.

So you mentioned talking about like prioritization because you need to decide on what order you're shifting your data into the cloud. I'm wondering, how do you prioritize? Is it like the high value data first because that's the most important, or the low value data because it's less risky? Or, or do you do it by team? Or how do you think about this? How do you prioritize?

So that's a great question. So how we have been doing it is that we are, at the same time that we're doing our data transformation, we're also doing a transformation to be a better company. We're doing a lot of initiatives we're working on to be better, to sell more, to treat our customers better. So all of those new initiatives, what we're doing is that we're thinking these initiatives are going to need data. For one example is we're making our website better. Well, the website needs this type of data, all these pieces. So let's bring those pieces to the cloud. So when we create this new website, it's going to use data from cloud instead of using data from premise. So we prioritize by the data needed for the new stuff that we're bringing in. We are doing it all with data from cloud and then we start thinking about what is the data that we use the most that is used in most systems, in most cases that it's like very really important to us to report on our business. That's the data that comes in as well. So we are bringing in the first group of initiatives that we had. We saw what data we needed, then we saw, okay, well, what's the biggest database that we're using the most? Our retirement database. We brought that in. And then these next set of, for the next couple of years we're looking at, okay, what are the initiatives that we're going to be working on on the next couple of years? What data do they need? What data do we don't have yet in cloud that we're going to need? And then we bring that in and really the data that is used the least or by the least systems, least people, these programs, that's the one that comes at the end. In a perfect world, we want everything in cloud and that's where we're heading. But some things are going to take a little bit longer and we have to be okay. It's a journey, it's not going to happen in a day. So you have to be patient and you have to keep going and keep at it to make it happen.

That's a very good point. I've noticed that. Well, basically everywhere I've worked, management tends to have like, like a short amount of patience for these really long technical projects, unless they see some kind of benefit early on. So is there anywhere where you think you've had like an easy win or you've been able to demonstrate some value from this data transformation program kind of part way through rather than having to wait till the end?

Yeah, no. So we have some incremental value on the way. You're completely right. You have to show some value added because if not, it's, it's just like putting a lot of money into it and then you don't see any results. That never goes well. So what we're doing is that as we're building this foundation for us of bringing this data in, we're starting like we have already a couple of machine learning models that we use. Just use like all our data is already in cloud. There's other initiatives that have happened. Like we had did some customer mastering and that data, it's already in cloud. The mastering that we produced. There's a couple of other big initiatives that they were related with our website and interactions with our customers and that was all the data needed for that was also in cloud. So we've had some early wins but we keep going as we go and have some more wins. So the idea is that as we are creating all these initiatives, that's why we prioritize that way so we can start getting value added by having this data already in the cloud.

And so with these sort of big technical projects, it can sometimes feel like it's a sort of back end thing that's a bit removed from the customers. I'm just wondering what's the impact on your customers been so far?

So our customers like they don't need to know or they should not care about where we have our data. What they want is having good data, right? They have, they want to have it on time, they want to be able to see the data when they need it and they, they want to have a better, better digital assets or interactions with us. Right? So that's where they have been seeing the results on what we're doing. They will not know why, but suddenly the website works faster or for example, suddenly the calls are being routed in a better way and they really don't need to know exactly how the data is going from point A to point B and why it's taking longer or shorter times. But they see the benefit there. So as I was saying at the very beginning, by what we do and by going through data transformation, by having applications of machine learning and AI, what we do is really improve our customer service and then doing that, then we also are able to grow our business and also keep our clients and they keep, keep them happy. Right. And reduce costs for us so we can pass that along as well. So it's all, all good. You see, there's no downside other than that it takes time and a lot of work. I think it's a great thing when companies go through these data transformations. I hear again and again everybody's doing it. It's kind of like something that we have to do at this point. We cannot just stay in wait. Right. We have to do everything we can to be in a better place. And that's what we're doing.

Absolutely. So I'm curious on what the timescale is beyond. It's a long time like, so when did this program start and when do you think you'll be done?

I think it started a couple years ago and we're hoping that it's going to be done in a couple years. So I'm thinking it's going to take about four years or so. There are some pieces that started that they're like, they're starting as we go and then they will end up later. But I think that's more or less the time frame from beginning to end. So it's very cool transformation. I think 19 is when it started and then should be done by the end of 2023 or half of 2024.

If everything goes to plan. The end of 2023, more realistically a bit later. Okay, I'd like to talk a little bit about the the technology's using. So obviously you're adopting some cloud tools. Has your technology stack changed at all beyond that as part of this transformation?

Yeah, definitely. So we were using some cloud already a couple years ago, but not that much. So we were developing our models for machine learning and we were using tools like Domino and we were using Hadoop and BitBucket. Right now we moved to aws. So that's the cloud technology that we're using. We're working on their SageMaker and environment for machine learning development. So we're using now SageMaker and we're using Redshift and S3 buckets and those pieces there. But we also were using BitBucket. So our tool stack, it did change a little bit. The idea is that as we move more data into cloud, it's going to be a lot easier for us to run the models that we're running and more and more running them in real time. While now we do batch with batch one. So it has changed. We had to develop a new infrastructure for us because as you can imagine, like every company also has to look into their security and what works and whatnot. So you have to do a mix of what it's already out there and then you put your own guardrails and follow those good practices that you have for your company. So we integrated those and we're super excited because we finished our platform and now we're developing there and more and more we're going to be able to be more efficient in my group. So it's all really very exciting times for us.

Wonderful. And so because this is such a huge effort, which sort of other teams have been involved in this, beyond just your sort of analytics and machine learning teams.

So the data transformation has been a huge effort that the whole company has been involved. You have from our leadership on the business side and on the IT side, our CTO has been instrumental on this. And if you think about like all the teams with IT that they're needed, you need the production teams, you need the strategy teams, you need the DevOps teams like architecture, engineering, like, there's a lot of teams that need to work on these data transformation. Some are going to work on how we make the infrastructure, others are going to work on how we bring the data. The data governance, data quality and data science teams are important here. The business and then business analytics teams are important as well because they have to set up requirements of what they need in this environment to be able to do bi and have reporting the business. They need to be really involved and supporting because all their processes that now they get their data from like servers on prem, now they're going to be getting their data from cloud and that opens a lot of possibilities, but also a lot of challenges on paying like making sure that they're on board so they can tell us exactly, like, oh, this process is getting its data from this place. Let's make sure that when we move to cloud, we can keep doing this process and we point it to the right place. So that's the beauty and the challenge of a data transformation is that you need everyone. You cannot just do it on your own and in silos because then it does not work as well. So it's a lot of coordination, a lot of collaboration and a lot of compromises. Right. That you have to make as well. You have to start really thinking about what others need and not what you need to make this work and then figure out something in between. So it's a lot of different teams working on it, but definitely worth it.

Okay, so all this sort of collaboration between lots of teams, I know it is often really, really hard stuff. So I'm wondering, how do you manage all these teams having to communicate with each other and collaborate?

So you set some processes. You said the leadership has to be aligned. So it starts there with the leadership really being on board, having our CEO and having our CIO and our CTO all thinking in the same way and thinking where we want to go. That's one piece. The other piece is that when you start getting more tactical on how we get things done, we have tons of meetings between several teams. So, for example, for figuring out what data we're going to bring in. And I'm working very closely to that one. I organize a meeting where I invite architecture, I invite engineering, and I invite the business, the program management office, and also our data and analytics team. So that way we understand, okay, what are the requirements of data from these business, owners of these processes, and then what is the data that it's already in place. So we talk to architecture, engineering, and how are we going to bring it in? So we have to talk to them as well. And then governance really helps us out on, okay, how are we going to govern this data, how are we going to curate, what are we going to be looking at when we're thinking quality and what is going to be the right source? It's not just bringing the data and dump it there. You have to figure out if you're, if you want to say, like the name, where do we bring name? From this database, from this database, from this database, which one is the right name? So we have to do some mastering there. So there's a lot of collaboration between these teams. And what we do is that we, we meet regularly and we break it in pieces, right? They say, how do you eat an elephant? A bite at a time? Well, how do you do a data transformation a few data items at a time? You just start like putting in little, little pieces and moving those pieces and making sure that everything that you do follows that, that same purpose and you're doing in the same way. So it's easier to get to where you want to go.

I imagine with a program of this scale, something must have gone wrong somewhere. So I'm just wondering what have you found that has been challenging or is there anything that, that you wish you'd known at the start?

So I think that something that is very challenging and we have learned on the way is that you need really good requirements at the very beginning of everything. You have to make sure that you don't skip any pieces when you're putting your requirements together. So, for example, if you're going to have a project where you're going to bring data from many different other places, if you forget a couple of pieces at the time that you need them and they're not there, it's a lot harder to bring those pieces in. It's a lot easier when you plan ahead and you say, okay, these are the pieces I need, these are the transformations I need to do, and this is where I'm going to take it from point A to point B to point C. And this is the final place where it's going to leave. I'm going to curate it this way. It's a lot easier than bringing 80% of it and this like, oh, the 20%. Oh, and we need this other piece and it's not as efficient. So that is one thing that I think is challenging and it really makes sense to spend the time before you start moving data to really have those clear requirements. That's one piece. Another challenge is that you have to keep doing what you're doing and making room for new. So you have to be making sure that you're not doing your everyday job right. And at the same time you have to put a lot of emphasis on the new stuff. So that means more work and means a lot more effort. Totally worth it. But you have to be careful of how you do it so you don't not do your BAU work and business as usual. And at the same time you're building something new and then at what point you move from the old to the new, you have to really test well. If you don't test well and can you imagine events you don't have the old and the new doesn't work, that would be really, really bad. So that's something that I think that we all learn the hard way at some point when we think we're going to go into production on something and it doesn't work as we thought it was going to be because we missed a couple of pieces. So it's always good that you have that plan B of okay, before I go into production, I'm going to test it and make sure that it's going to work. And then you, you keep both at it for a little bit and then you, you cancel the old. So those are things that they're challenges, but they're definitely things that we have to think about and always think about a plan A, plan B and plan C just in case something go as planned. Because as you plan, they plan for the worst and expect the best, I say or something like that. The saying goes, I'm not sure, but you better plan for everything.

Okay, yeah, so that does seem really important. The idea of like trying to avoid introducing new bugs into processes just as you try as you move data around. So I'm curious if you have any more to say on how you go about testing things.

So yeah, so we have a really good programs to test. For example, I can talk a little bit more on the machine learning side. We make sure that we test on our own environment. We have a research environment that is a prod environment because we use prod data training, but we're in a development environment at the same time. So we do all our testing, we check that our models are working, we make sure that the output that we're getting is what we're expecting. And once from there, then we take it to all the environments. So we go from there, we go to dev mode, then we move it to test and we move it to model, then we move it to prod. So in all those jumps we're checking and rechecking that everything works, that we're not affecting any other processes or any other pieces. Something also that we do to for testing is that our production team has a production process that you have to go through it. And as we're moving through the environments, they check their scan, they make sure that if something breaks, how can it be fixed? And then by the time that it's in production, we're pretty comfortable that what we did is what we're expecting and that there's not going to be any issues. And we have always the plan B. If there are some issues, what is the way of solving those issues. We always have that also ready to go in case something would go wrong.

As well as having this sort of multi layer testing thing, you've got ways of like diagnosing the problems and having backup plans for like what you do when you do.

Yeah. So we know, okay, what happens if suddenly we lose the whole data for a day or two. Well, we have, oh, we can use this backup. We can use these like there's always a plan B there of how to mitigate the issues that we may have. And depending the severity of the, of the problem or the importance, how many systems will be affected. Then we have backup systems. If something fails, then the backup comes in and so we make sure that we're always is in a good place. So that's something that companies do, including ours, to make sure that we're meeting gating any issue that it may happen. Right. So that way, like if you imagine if companies didn't do this, then you would not be able to do anything. Right. Like suddenly your bank is down and you cannot do anything that doesn't work for very long.

Yeah. Hard to make money when all your systems are down. Okay, I'd like to talk a little bit about skills. So it seems like because everything's changing quite fast within your team and more generally within your organization, how has that changed the skill set that you look for in your team?

On the skill set what we're looking for is really for data scientists and people that they're willing to learn. Because things are going to keep changing. What it was like years ago, it was some programming language, then we changed, then Python became like the one that we're using. But then if you're using, if you're in cloud, need to know a little bit about how to deploy in cloud. And then depending on. So everything changes, right? The tool stack may change again. So when I'm looking for people for my team, I'm not looking just for what they know, but how good they are to learn, how willing they are to learn. Because that's the most important piece that I see for data scientists, at least for machine learning, for AI. You have to be ready for change. We may have Salesforce right now as a CRM, but who knows, maybe in two years we change to something else. So you have to be ready to think in a very open way about how can we integrate the output of our models if we change systems or if we bring a different application that we cannot. We don't even know that exists. Right. So maybe in two or three years that completely changes. So we have to be ready for that. In the skill size, I would say for my team, I'm always looking for strong sense of statistics and math, that understanding of science, of how you think in a. I'm going to have a hypothesis and then I'm going to prove it and then I'm going to do these, like having a very organized mind of how you're going to approach a problem to solve it. I think that's very important. And languages, we can learn them. New software, we can learn it. But what we, what is hard to teach is the ability to learn. And that's what I'm always looking for.

Okay, Certainly that point that technology changes fast and the bits of software you can be using will change every few years, that really resonates with me. But yeah, I like the idea that you always need to be willing to learn new things. So on that note, actually, when you do find that you've got a skills gap within your team, have you been training people internally or do you hire the skills like from new people outside your organization?

So we have, we have done it both ways. Like sometimes I get people with like skills that they bring with them. Other times I also, our team is very supportive about training to learn new skills. So one of my team members, for example, he's very passionate about natural language processing and he had, we have provided a lot of training on that side and he has learned a lot on the job as he goes. Right. And in other cases, one of my data scientists, a statistician, brought a lot of knowledge on that side on statistics. So I think that for machine learning and AI team is very important, or at least I find it very important, that there's different backgrounds, that one of the beauties of data science that you can come from being a physicist or being a statistician or a computer scientist, there's a lot of different backgrounds where how you get there. And for us, it's amazing when that happens because that way they bring different skills that they can share and teach to the team as well. So something that we do and we are very purposeful about is that we have a lot of sessions about sharing so they can help each other and learn from each other. And to have a data science successful team, you need to be able to do that because nobody is going to come in with all the skills. There's no way that's going to happen. And even in your own team, not all of them are going to have all the skills. So you need to have someone that has very strong skills in one area and others that have very strong skills in other and then they share and teach each other and help each other. That's something that I value the most. I know with Data Camp that they go in and every now and then one is going to be looking into deep learning, another one is going to be looking into maybe pyspark. So depending on what they want to learn, they're going to be moving in different directions. And it depends what they're specializing on as well at the moment, what they're going to have to learn.

I love that your team's using Data Camp for continuous learning and improving their skills. That's wonderful. So you've talked about how your team, like they need to be good at translating technical problems into things that business people can understand. And the importance of having a learning mindset and the importance of understanding statistics and hypothesis testing. Is there anything else that you think makes people in your team successful?

So I think the creativity is super important because not everything goes as we want it to go. And having that positive attitude about finding a solution, we don't have the chance of saying, oh, no, it cannot be done. We are more about how do we make it work. The data is not in the perfect place. Well, we make it work. We have to adapt to these way of doing things because it's going to keep our data safe. Well, we adapt and we make it work. So in my team, for me it's very, very important. Somebody that when they see a problem, they're creative and finding a solution and not giving up, just like figuring it out, how to get that solution to me is very valuable. And it happens more than one thing. You go to school and they give you the perfect data set and they say, build this beautiful model and it always works, right? So you're like, oh yeah, I tried these five different techniques and all work really beautiful. You go to the real world and it's like, well, how do, where do I start? The data is really, really weird. Well, all those pieces, being creative, it's going to get you there. So to me, creativity and positive attitude, that's really what is going to make it happen.

Absolutely. I like that. Okay, so just finishing up. Is there anything that you're really excited about in the world of machine learning and AI at the moment?

I have to say that I love it all. Something that it gets me really excited on this world is really the possibilities of making change. I love models that when you create them and you have an output, that output is used in a way that the customer doesn't even need to know. In this case our business. Customers like our business side of the house. You make their life easier in many ways without having to worry about it. It's like automatic, that AI piece of doing prescriptive stuff and taking decisions on the go. That makes it to me super exciting. And being able to use real time data and running models, real time. I think that's something that keeps me very excited every day and I'm looking forward and working on as much as I can.

Wonderful. Yeah. So I think AI and being able to do data driven decision making automatically. That sounds fantastic. And yeah, real time analytics also wonderful stuff. So do you have any final advice for other companies trying to get started with their data transformation program?

Yeah, I would say that don't think about how complicated or how big it is is, but what you're going to get from it. So I think that my biggest piece of advice is that it's not easy and it's long and time flies when you're having fun. So just enjoy the journey and make it happen. I think that that's what I would say about data transformation and for machine learning and AI. I would just say there's so much we can do and such a big difference you can do wherever you are. It really doesn't matter in what industry you are or what type of business. There's always a way to help people and to help everybody else. To make their life easier that we can use. So that's like, if that's what you care about, it's a great field to.

Be in, making other people's lives easier. That sounds wonderful. That's great. All right. Thank you very much, Vanessa, for your time. That was really, really informative. Thank you very much.

Thank you so much. And thanks for having me. And best of luck to everybody that they're building a career in data science and machine learning and data transformation. Super fun thing to do. You've been listening to Data Framed, a podcast by Data Camp. Keep connected with us by subscribing to the show in your favorite podcast player. Please give us a rating, leave a comment, and share episodes you love. That helps us keep delivering insights into all things data. Thanks for listening. Until next time.