HyperScience was proud to be an official streaming partner of NeurIPS 2019, the largest AI and ML conference in the world. On the heels of the event, we caught up with Momchil Rusinov, one of our Directors of Engineering, for his thoughts on the conference and key trends in the world of ML.
Who are you and what do you do at HyperScience?
I am a Director of Engineering in the Sofia office, responsible for our Machine Learning and DevOps teams. My goal is to ensure that our teams are happy and productive, as well as to remove any friction and minimize uncertainty and ambiguity in requirements. I help with process-related issues, people management, and last but not least, making sure that everyone on the team has the opportunity to learn new skills and grow as a person and professional.
Can you tell me more about HyperScience’s NeurIPS 2019 Meet-Ups in Sofia?
The Conference and Workshop on Neural Information Processing Systems (abbreviated as NeurIPS and formerly NIPS) is the world’s largest Machine Learning and Computational Neuroscience conference held every December. The demand to attend is high, and we were fortunate to send several ML engineers to the event, but in 2019, we wanted to do even more to share the insights and experience with the local ML community in Sofia. As a result, we decided to organize a full week of meet-ups, during which we streamed part of the content from NeurIPS in our office and invited ML researchers and engineers to come watch and discuss with us! It was the first event of its kind in Sofia, and we’re already planning for additional follow-ups to discuss different topics we weren’t able to cover during the week.
What were some of your biggest takeaways from the conference this year as well as the discussions during the Sofia meet-ups?
There are significant efforts underway to provide tools for making ML a more mature discipline. There was a lot of focus on academic reproducibility, including in the conference’s processes. There were also a lot of papers on uncertainty estimation, explainability and overcoming hidden biases. In Sofia, the discussions were mostly centered around reinforcement learning and behavioral modeling.
What key trends in Machine Learning are you and your team following?
One of the key topics we are following is the development of the Transformer models. We are constantly trying to improve our performance and also the efficiency of our ML models when it comes to training and inference.
What are you and your team solving for in 2020?
On the day-to-day, we’re constantly working to improve the performance of our existing ML components in a scalable way. At the same time, we’re investing in developing new Deep Learning techniques that will enable us to handle more and more diverse documents and inputs in order to open the door for brand new features and functionality.
What advice do you have for someone who is looking to get started in ML/engineering? What resources would you recommend?
Since joining HyperScience, I’ve seen an uptake in Machine Learning driven by computational power and advancements in algorithms. I find that the best way to learn, especially in a leading-edge field where primary research is happening and things are evolving quickly, is to tackle problems head-on and dig deep to discover how they are solved. This is something we do regularly in Sofia via our own hosted meet-ups or by participating in other community events. It’s important to get together as a community, discuss what’s new and emerging, share best practices and try to tackle new challenges together. Trying competitions like the ones offered by Kaggle can also teach you a lot!
What’s your favorite part about working at HyperScience?
Without a doubt, my favorite part about working at HyperScience are the people and the interactions I have with them every day. It is really inspiring to work with such motivated, ambitious and capable experts across all fields, whether that’s engineering, product, sales, etc.. All this combined and crystalized in a small-but-growing company of 120+ people gives you the unique opportunity to learn and develop in so many different directions!
Copy: Momchil Rusinov