Daniil Polykovskiy
Member Of Technical Staff at MakerMaker.AI
Montreal, Quebec, Canada
Summary
🤩
Rockstar🎓
Top SchoolDaniil Polykovskiy is a machine learning and product leader with 11 years of experience building AI-driven drug discovery platforms and teams from research prototypes to customer-facing SaaS. As a former Sr. Director of Technology and IT Director at Insilico Medicine, he launched Chemistry42—the world’s first generative AI platform for molecular design—and unified product practices to accelerate delivery and adoption. He combines deep research credentials (NeurIPS, Nature Biotechnology publications and a PhD-level background) with hands-on contributions to open benchmarking tooling like MOSES, where he implemented core molecular metrics used to evaluate generative models. Now building toward AGI at MakerMaker.AI in Montreal, Daniil excels at aligning roadmaps with real-world discovery needs and turning complex ML research into practical, revenue-generating products. Colleagues describe him as a pragmatic innovator who values teamwork, creativity, and continuous learning, with a rare track record of taking generated molecules from model to in vitro validation.
11 years of coding experience
8 years of employment as a software developer
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at Higher School of EconomicsLomonosov Moscow State University
Russian, English, French
Github Skills (12)
generative-model10
rdkit10
machine-learning10
benchmark10
drug-discovery10
benchmarking10
python10
numpy9
data-analysis9
pytorch8
scikit8
scikit-learn8
Programming languages (4)
Github contributions (5)
molecularsets/moses
Jun 2018 - Oct 2020
Molecular Sets (MOSES): A Benchmarking Platform for Molecular Generation Models
Role in this project:
Data Scientist
Data ScientistContributions:46 commits, 34 PRs, 88 pushes in 2 years 4 months
Contributions summary:Daniil's commits primarily focused on initializing and developing metrics for evaluating molecular generation models. They implemented various metrics related to molecular properties, including logP, SA, and QED scores, as well as methods for calculating similarity and diversity among generated molecules. Their work included the implementation of the Frechet ChemNet Distance (FCD) and the integration of existing tools such as the Natural Product-likeness score (NP-score). These contributions directly support the benchmarking platform's core functionality.
benchmarkingmachine-learninggenerative-modelsdrug-discoverybenchmark
insilicomedicine/DD-VAE
Mar 2020 - Sep 2020
Deterministic Decoding for Discrete Data in Variational Autoencoders
Contributions:2 commits, 1 push, 1 branch in 6 months
variational-autoencoder
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.