I think I kind of unique story because I actually number graduated college, but probably the biggest is permission for me in my career. Trajectory was getting involved with the Oscar, your ship club of my school, because I had a very interesting exciting And because of that, I was like, I don't wanna go to tack it up going to start ups. And I also took the sheen learning courses, and I thought that was, like, the coolest thing ever on does at a time like the carving was getting really powerful and stuff. Um, so I can I was like, OK, I definitely want to get into a I definitely want to get into start ups. Ended up getting an internship after my senior year at this awesome startup called Data Hi to the super high growth one. I was data science in turn there. Um, you know, they get the position was really talk to you. I mean, you have upper grad degree. You know, I looked unlike again 95% of you live in a masters, so it was like, Oh, okay. So I really, really need Teoh. Do whatever I can to show them that I really want work there and that, you know, to be capable employees. So I like spent by two months trade, like going through online courses. You know, making sure I knew all the data science library research. A bunch of stuff built the company said a cover letter to a bunch of employees Their c o on the most sense, the CEO. That was when I got my first interview. Yes. Yeah, just like trapped a lot. Got the whole and then on the job there, I just, you know, was immersed by working a lot of different companies on data science problems on, um, I learned a lot on the roll. Uh, but I got a little bit of likes. Don't use the word sick. It was like a nice job was really falling. Oppa travel. I got to see it much. A. I use a different companies, but, you know, I was going to try to be an entrepreneur on, um, I wanted to get experience of actually building a product being a part of early stage company. So I one of my friends was working on a chat about soda and there are looking to hire their first engineer when I did a lot of natural language processing itself, But my first job, um, and I decided to just take the risk and go for it. And, you know, from Data Scientists software engineer, which was actually a much bigger change. I thought it would be actually like you both code. But the self you need to think about and actually do is kind of like, completely different. Um, it was a big adjustment, but I adjusted to it, uh, getting acquired. You know, I'm months into the working there and, uh, been with themselves.

So my role is, you know, software engineer, you know, basis on my computer all day. Uh, so I don't write at the travel that much? I couldn't work remotely, anywhere. I want this off, which is nice, but you have pretty pandemic at a office. We worked. The whole team was based out of there with me every day. Um, in regards. So, like responsibilities, I have It's manly, you know? Just kind of the new products. Um, making architecture decisions of how we want to build this so it could be robust on. We need to make changes in the future will be easy. Todo You know what I do a bit more than this coating. But I helped make sure that our servers are set up properly. That is a bigger change is that we got more customers that it can scale. Well, you know, the server or crash. Um, you know, I also do what, working the machine learning there. So, um, get a chance to and through the milky kind of. But but overall, it's just like whenever we need to get done, you know, to actually build the product. Uh, you know, just work and work on that on that entails, you know, mainly first, defining what needs to get done and coming up with a strategy of how we want to do this. You have multiple people or elaborate on this together on then we distribute tasks to certain people. Um, immediately like, one person would be like the lead engineer for a project, so they would mainly just focus on getting that dog. Yeah. I also helped by customer support something like a technical block, although pygmy. Okay, let's you know what was going on. I will do it. Um, the guards doing hours, uh, it's I think it's back. It's all to do the last step. It's supposed to 40 hours of eight hours. Yeah, working remotely. Now it's more flexible. Um, but generally there's not really a set. Our, like, sort it. You're getting yourself done. That's what counts. Really? Um, you care that much. I mean, any care, but it's not like you need to be sitting at your desk from here to here. As long as they're your delivery. Good results

eso being like a I driven product. Most the AI libraries are. So we manly used, I thought for our back toad. Um, like the machine learning is already python. But when you're when you're building a A, you know, an AI product that's Pash is building the machine learning models. Um, you know, you have to build, like, a web framework way. Use this pipe on library called Django, which is very popular. Built the back end software, um, in regards. So, like the algorithms or used before the I working about a lot, but we tried a bunch of different soft. You know, what we're trying to do is like, So we're, uh you know, Bach company, right? You know, someone sends their bottom at stage of me trying to intelligent response on what the eyes that we taken the incoming tax and we tried to make a prediction. What? They're actually I look the intent of the messages. So in a data science terms, that would be a multi class classification problem because we have, like, 60 different things. You want to try to predict the next. So we assigned a probability for each other modeled over the top. What is because we think the users saying this on way China on different models, including, like deep learning a tensorflow care us. And one day I ended up working the best. There's a lot out there, but there's actually impotent amount possible models you can use even like how much you can change it. No, not working different layers. And, um, we're trying a bunch for the one that worked the best. Was this one call fast text read in my face like, yeah, surprisingly, is actually like this simpler version, um, applying to the text of models out there that exists. I don't want to get too much specifics of it, but, uh, you know, generally, I mainly work in actually software engineering, so build any softened pipes on. You know, we have, like, a post grass database, uh, Canada database, storing soft taking input data. But when people look at something on the web site doing some action