Michael Gimelfarb
Postdoctoral Research Fellow at Department of Computer Science, University of Toronto
Toronto, Ontario, Canada
Summary
👤
Senior🎓
Top SchoolMichael Gimelfarb is a Postdoctoral Research Fellow in Toronto specializing in reinforcement learning, transfer learning, and offline RL with eight years of experience bridging theory and applied systems. He develops scalable gradient-based planning and generative-model-driven sequential planning methods, and has a track record of publishing at NeurIPS, ICLR, UAI and AAAI. His work includes practical engineering: building Python toolchains that auto-generate OpenAI Gym environments from PDDL, managing CI for a planning competition, and prototyping research code in TensorFlow and JAX at DeepMind. Comfortable moving between optimization (Gurobi), robotics, and large-scale empirical evaluation, he focuses on making planning robust and explainable in high-dimensional settings. An economist-turned-ML researcher by training, he blends operations research rigor with hands-on software development to turn complex RL ideas into reproducible experiments and usable tooling.
8 years of coding experience
7 years of employment as a software developer
Doctor of Philosophy - PhD, Operations Research, Doctor of Philosophy - PhD, Operations Research at University of TorontoBachelor’s Degree, Finance, Bachelor’s Degree, Finance at Schulich School of Business - York University
Github Skills (44)
simulation10
sgd10
visualization10
controller10
simulator10
gymnasium10
visualizer10
openai-gym10
benchmarking10
reinforcement-learning10
backpropagation10
evaluation-framework10
mdp9
nonlinear-dynamics9
keras9
Programming languages (4)
Github contributions (5)
pyrddlgym-project/pyRDDLGym
Sep 2022 - Mar 2023
A toolkit for auto-generation of OpenAI Gym environments from RDDL description files.
Contributions:4 releases, 4 reviews, 714 commits in 6 months
openai-gymgym-environmentsmodel-basedrddlreinforcement-learning
mike-gimelfarb/deep-successor-features-for-transfer
Jan 2021 - May 2021
A reusable framework for successor features for transfer in deep reinforcement learning using keras.
Contributions:38 commits, 4 PRs, 34 pushes in 3 months
deep-reinforcement-learningkerassuccessor-featuressuccessor-representationreinforcement-learning
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.