Why I hate quitting Deepomatic
Every startup has many firsts during its early years. We have known our share of exciting firsts: first employee, first fundraising, first signed contract, first office outside of the incubator, first discussion with France’s Prime Minister, first seminar abroad, first profitable month,… This month, I am the first employee to resign. Although it might not be as exciting for Deepomatic, it gives me the occasion to reflect on the year I spent here.
Near the end of my PhD thesis in logics, I read that Deepmind had created a program which could learn to play atari games from pixels. At that moment, I became convinced that artificial intelligence would play a huge role in the following years and I wanted to be a part of it. That is why I applied to join Deepomatic, to learn as much as I could about machine learning, and to use this knowledge on challenging problems.
I was not disappointed. Vincent — our CTO who did his PhD in computer vision — shared his knowledge with me, pointing to important resources and explaining the most tricky parts. I had the occasion to work on many different problems using various machine learning techniques: ensembles methods to improve accuracy, distillation, autoencoders, contrastive loss, weakly-supervised, semi-supervised and active learning, topic modelling, personality clustering on social networks,… I had both the occasion to read and implement recent papers, and do original research. Many among us were interested in machine learning and artificial intelligence advances, even outside the scope of Deepomatic. During the Tuesday research meetings, we had many stimulating discussions about scientific papers. Papers we wanted to implement, theoretical papers, or just papers we found interesting.
While this has been an interesting year, I am mostly interested in very long term research, my dream being to contribute to the first artificial general intelligence (AGI). This is not the purpose of Deepomatic, so I applied to Vicarious (a company aiming to build a unified algorithmic architecture to achieve human-level intelligence) and I was accepted. Because Vicarious is a high profile and very selective company, this attests that Deepomatic was a great place to grow in competence both in terms of code and machine learning. Even more knowing that I had never programmed in python before joining Deepomatic and that my machine learning knowledge was minimal.
So, Deepomatic wants to hire a new data scientist. They will also need to find a replacement for my other job there. Do you think you have what it takes to be our new Beer Captain? Every Thursday afternoon, we have a team meeting. We discuss the latest business news (new contracts, prospects), our progress on current objectives (with demos), our next objectives, and any subject people want to bring up. And, to combine work and pleasure, we taste craft beers. As the next Beer Captain, you will have to remember everyone’s preferences (small tip: get white beer for Zoé, and IPA for David). The two positions are obviously not linked, you can apply to be a data scientist without being Beer Captain material.
This is one of the small pleasures which, because we shared it together, brought the team closer. So did the chocolate croissant of monday morning, the footsie games, doing sport together (from jogging to tennis), the team seminar we had in Morocco, the Thursday meals Nolwenn cooks. We also have every quarter a “King for the Week”: an employee gets to decide the priorities for a week. Kevin got the lucky charm in Epiphany’s king cake and we focused on small projects we had delayed for too long, especially concerning our new office (for instance, we can now open the building door with our smartphone instead of bothering others with the intercom). Alexis was the next King for the Week and we focused on getting familiar with each other’s code and tools. More generally, all those fun examples are the manifestation of the genuine attention of our founders, towards us, our opinions and our well-being.
While my dreams are bringing me to other places, I leave Deepomatic with a heavy heart, because I had a great year there, and I will miss those colleagues who became friends.
If you are interested in the newest developments in deep learning and/or want to be a part of Paris’ coolest start-up, be sure to visit this page :)!
