I learned a ton in the ten weeks I spent there–and not only about Machine Learning, which I’ll get to in a moment. My main motivation for pursuing the internship was a desire to further my knowledge of data science and gain a better understanding of what it’s like to work full time in this industry. Now that the ten weeks is up, I can wholeheartedly confirm that I managed to accomplish both goals.
Prior to the internship, I had taught myself most of what I knew by reading, doing online courses and working with simpler data on Kaggle.com. If you add up the amount of time I devoted to my development, it wouldn’t have exceeded 7-10 hours a week. As an intern at deepsense.ai, I had the opportunity to develop 40 hours a week in a very interesting and effective way.
Along with another intern, I was assigned for the entire 10 weeks to a project that one of the teams at deepsense.ai was working on. The project involved developing a predictive modeling model for a client in the advertising distribution industry in applications for mobile devices.
We spent the first week learning about the data we had access to and understanding the business problem to be solved and the technical limitations to deal with as we worked out the solution. The main complications in our project were „Time Series” and „sparsely linked data with user id.” It would make for good material for the film “Inception 2.”
The second complication was equally inconvenient. Imagine how Netflix recommends movies for you to watch. They base their suggestions mainly on movies you’ve watched in the past, and about other movies that users have watched, who have also watched the same movies as you. In our situation, there was no such feature as a “user” so we could not look at ads that one had seen in the past or ads that people similar to you had seen.
In later weeks we had numerous meetings (at least two per week) with our brain box of a mentor, Mateusz, who taught us a ton. I sincerely thank him for that.
During our meetings we would tell Mateusz how things were going with the jobs he gave us. Once we had talked through a problem, he would give us a new one to solve. It was during these meetings that I learned the most. Beyond updates on the project, we had the opportunity to talk about related topics we didn’t fully understand, analyze our thinking and ask about other concepts we were curious about. Matt was always glad to answer our questions and engage in new discussions. Because he knew so much, our meetings were really educational. They were also among my favorite aspects of the internship.
Beyond the meetings, I learned a lot thanks to just spending time in deepsense.ai’s Warsaw office. Just imagine being in a building where 70 data scientists work, and each of them has more experience than you! (I would really stress that deepsense is bursting at the seams with intelligent people). If I was reading an article on Machine Learning and came across something I didn’t understand and couldn’t get a satisfactory answer to with a quick Google search, all I’d have to do is walk 20 meters to our fun room and there would always be a few people there who could sort it out for me. It was great! We could also ask questions on the company chat, and get the answer within a minute.
Lunches together, a quick game of foosball and video games in the fun room were also a very nice part of the internship. Every day around 12, everyone orders in lunch (compliments of the company, of course). When the food arrived, we’d make our way down to the cafeteria and take a break from work and talk to friends – not always about data science…
A quick game of foosball was also a very popular way to take a break. A special chat channel was even created to organize the quick meetings. All you had to do was put the question out there: “anyone up for a match? We are looking for 2 more people.” 30 seconds later, 4 people would be in the funroom spinning their wrists for ten minutes.
I am very happy with the internship in every respect and plan to return to deepsense.ai next year when I graduate.