An ML intern's second day: senior engineers Connor and Athan spend the week cleaning raw data while he reads an arXiv paper on entity embeddings, implements it in Docker, and prepares for model training. He learns why meeting with clients first matters.
Chapters
Hey Mum, what's your favorite voice assistant? Voice assistant? Yeah, Siri, Alexa. Oh, I like Alexa. Alexa? Yeah. Really? How come? Because she's organizing your life at the moment. What about Siri? Siri's good. Yeah? Yeah. What'd she help you with? She helps me with, yeah, timing, cooking, reminding me. So yeah. Mum's dropping us to the train station. Me and my brother. What's up, bro? Sup, Millsie? We're heading off. He's heading off to work. I'm heading off to day 2 of my internship at Max Kelson, which I'm really looking forward to. I'm not entirely sure what I'm going to be getting up to today, but once I find out more, we'll check back in. See you, Mom. Love you. Morning. My hands are so full right now. I got coffee, lunchbox, because I don't like buying food when I'm, when I'm out place. I like eating my own food. I'm in training for Spartan Race now, so I have some pretty high standard nutritional requirements. So I make my own food, backpack, laptop, whatnot. I'll tell you a little story. I'm just about to walk into the building where I'm doing the internship at, and the people at the cafe I went to, I walked in and they were really upset. I was like, oh, what's going on? It was just a little cafe that I hadn't been before because I like trying new places. And I walked in, I'm like, what's going on, guys? And they're like, we're being shut down. Um, so apparently Council came along that morning and told them that they can't use their storefront anymore because the building they were in were complaining about the tables and chairs being on the street side. So essentially they've got no space at all now. They're just a hole in a wall, and that really sucks. But they, they insisted on giving me this for free. I was more than happy to pay for it, but they gave it to me and said to have a good day. Like, they were upset, but they were still in good spirits. They knew, they knew whatever they were what they're going to do next, they're going to try their best. So guys, I wish you all the best. It sucks you're being shut down, but I know you'll get onto something new and just smash it. Thank you so much. Cheers. Alright, I'm watching this back. And don't do this. Don't do what I'm doing. Either computer or eat. Don't do them at the same time. Back to the video. So I kind of stuffed up. I got on the wrong train to get home, and now I, I was replying to messages on my phone and I ended up the complete opposite direction to where I should have been. But I figured while I wait for an Uber to come get me, what better time than to tell you what I got up to in the internship. So this morning I got in and it was about a week since I'd been there. Well, exactly a week. It was Thursday to Thursday, and the two guys I was working with had made some epic progress on the machine learning project that we were working on. They were doing a lot of data cleaning, and why did they have to do so much data cleaning? Well, when we got the data, or when they got to the data from the partner company, they didn't really look at it before they got it. So it was sort of like, here you go, here's all our data. See if you can run some machine learning stuff on it and find some insights for us. So that's probably a learn for the future. Before you go into any project to partner up, well, that's what we were talking about today, me and one of the machine learning engineers, Athan, who's an absolute wizard. Shout out to you, Athan. Before you partner on any project, actually, it's probably best to meet up and discuss in person what you guys are both looking out of it. And my Uber's just here, so we'll continue this in a second. Hey man, how are you? I'm just gonna put this in the back and I'll get in the front. Sandgate, please, my man. Thank you. Hey guys, this is my friend Zegwinda. You were just teaching me about investing. What's the number one tip for people here? How can I help them out? Uh, number one is Never, never put borrow money. Second is never put credit card money, only saving money. Only savings money. So no credit cards or no loans or anything. Yeah. Minimum you have to practice 1 year. 1 year. 1 year. Yeah. With anything, right? With not just investing, but you got to practice with anything. Anything you want to do. You have to keep in touch every day. Every day. The market. Every day. Well, thank you, man. Good to meet you. Have a good night. Thank you. Thank you. Okay, if you need any help— Thank you. Have a good night. See you later. Finally made it home. And where was I? We're talking about when you first meet up with a client, it would be ideal for you to both talk to each other and get on to terms with what it is you're actually working on. So, that was a great learn. I think it's Even though the stuff that we're working on today was quite difficult, I think that's the best thing to do. I mean, it's where— that's when you learn the most, right? When you're working on the hardest projects. What did we do after— oh, my puppy's saying hello. This is Seven. Say hello. Say hello to YouTube, Seven. You're famous now. What do we do after that? Well, they've been spending most of the project, Connor and Athan, the two machine learning engineers, cleaning the data and essentially getting it ready for putting it into a machine learning model. So that's the last little bit of it. What was I personally working on? Well, I read a paper on arXiv about entity embeddings, which I'm thinking they're similar to word embeddings, but they might be a little bit different. I'll link the paper below so you can check it out as well. And then I implemented the code from the paper onto our server, I believe it is, or Docker container. I put it into there and made sure it can run. So we're going to use similar code from that, from that paper to run a machine learning model on the data that we get once it's fully cleaned up. We've got a whole bunch of new data today that we had to add to the already processed database. So much fun, lots of learning, and I think I'll be back there next week. I should, should hear from them, from them soon. Exciting times. I'm going to go lift some weights. No, we ain't having none of that. I'm gonna go tell Dan I'm watching this video in 2048. We can't have it just ending on no hashtag. What are you thinking, Dan? Yeah, I know, but I didn't make any outro clips. I forgot. Hold on, I got an idea. What do you think? It'll do. Cue the music.
