Portrait of Krzysztof Maziarz

Krzysztof Maziarz

Principal Researcher

About

I’m a Principal Researcher at Microsoft Research AI for Science, where I’m applying Deep Learning to advance scientific discovery. Among other things, I’ve worked on generating drug-like molecules (opens in new tab) (code (opens in new tab)), predicting molecular properties (opens in new tab) (code (opens in new tab)), and retrosynthetic search (opens in new tab) (code (opens in new tab)). Notably, our pipeline of generative and predictive models of molecules are used in-house by Novartis to speed up drug design (opens in new tab). I’ve also made many other side contributions during my time at MSR, e.g. developed a hashing algorithm invariant to variable renaming which appeared at a programming languages conference (opens in new tab).

Before joining Microsoft, I obtained my Bachelor’s and Master’s degrees from Jagiellonian University in Cracow. During that time, I interned at several companies such as Google AI (opens in new tab), Tensorflight (opens in new tab) and Jane Street (opens in new tab). Most notably, I worked on training outrageously large sparse neural networks (opens in new tab), architecture search with evolution-RL hybrids (opens in new tab), learning multi-task network architectures (opens in new tab) and pooling multi-view features of buildings (opens in new tab).

My works were published at top ML conferences (NeurIPS, ICLR, ICML, AAAI) and domain-specific journals; for a full list, see my Google Scholar page (opens in new tab).

Outside of machine learning, I participate in competitive programming competitions, where my achievements include advancing to the finals of Facebook HackerCup (opens in new tab) (25 finalists worldwide), Distributed Google CodeJam (opens in new tab) (20 finalists) and Google HashCode (opens in new tab). I also ranked 14th and 34th on the ACM ICPC finals, and later served as a coach for the Jagiellonian University teams. I also can be found taking part in the finals of Algorithmic Engagements (opens in new tab) or participating in contests on several (opens in new tab) established (opens in new tab) competition (opens in new tab) websites (opens in new tab) where I rank top 0.2%. I also author problems for various competitions, including more than half of the tasks at the Polish ACM ICPC regionals (AMPPZ) in 2019 (opens in new tab), 2021 (opens in new tab) and 2022 (opens in new tab).