Hey, everyone at mentor students. It's nice to be talking to you. Thanks. Josina for a question. Um, so let me give you a quick introduction to my background. I started out as an engineer s. So I wanted to go into an engineering school, but ended up doing a masters and science from IIT Bombay. Then I got started with the PhD again in physical sciences on DWhite. I was in the physical sciences when I was getting my Ph. D. My interests diverged. I ended up taking B school courses at Booth Andi. Also, a lot of my research also went into the computational side off things. At the very end, I was doing more experimental and computational work, all right? And so is the national next step for somebody who was getting a PhD in science. I was looking for opportunities where I could make a lot of impact on one of those areas happened to be data science where I saw that data science is gonna have a large impact on multiple industries on that is what led me to get interested in data science. I was looking at different ways of breaking into this carry into this role. I found messages. Data science, boot camp. They help me quickly get acquainted with a lot of statistical, uh, methods on computational projects. Uh, on that basically got me my first role Where I was looking for a firm very Microsoft was the was declined. Uh, and so that sort of contract position. Then help me pivot towards more specific area. It was a supply chain, a supply chain data scientist and I now work as a supply chain data scientist at Chewy Eso. One thing to note is that data science being a new area, you have, you know, people from different industries. Uh, on and 10 years ago, data science did not exist as such. So the field is in some ways filled with people from physical sciences who have being able to contribute to data science on dso. I'm happy to be here

the top priority is to frame equations and objective functions to optimize the problem at hand. Uh, second priority again is very similar. It is the apply the right machine learning framework. Uh, so, for example, I have a specific project at hand where I'm supposed to optimize. Let's say the routing off orders, uh, then three idea is to be able to frame the right equations and objective functions, given the data at hand so that you are, uh, delivering on on them. And the third priority is continued and consistent improvement and model performance. So, being a machine learning model or be it you know, a simpler, uh, set off set off sort of techniques you'll be you have to be able to consist continually and consistently improve on those. The top three challenges are to be able to operate at scale. So sometimes you're dealing not just with terabytes of data but petabytes of data on bond. You have to be able to extract meaningful insights from them. The second big challenge is, uh, data cleaning, because often, really, world data is not clean, so getting the right data in place and the right format is important you have to make sure that the time that you take for look up is small on. The third challenge is constantly changing priorities as things evolve over time.in a format where, you know, you try to reduce the amount of time that it takes to read that data on day. Then be able Thio, you know, process incoming information. So sometimes you rely on information that does not change frequently, maybe once a week, but sometimes, but some of the other data is it. It changes every five minutes. Eso basically organizing data in a way that helped Thio improved model performance is a trick that that that's very helpful. And it is it is an old trick from there. People who are, you know, you had database administrators, but these days, you know that trick, I think, is sometimes underappreciated.

heavily with python eso, The data clients team developed Smalls in Python and then there is a software engineering team that then converted to Java. I also work extensively with Sequel. Uh, in my previous role, I had worked with some of the big data platforms. Uh, another important technique are another important, uh, sort of call it method, You know, our linear programming methods that are used extensively on and, you know, my team uses extensively Onda. For that you need solvers such as grow B or C plex. And so a lot of this work is directly derived from operations research on DSO. You know, reading scientific literature is also very important because you don't want to basically be able to. You don't want to build things from scratch if you can learn from others.