Rohan Potdar
Software Engineer at AMD
Lafayette, Indiana Metropolitan Area United States
Summary
🤩
Rockstar🎓
Top SchoolRohan Potdar is a software engineer focused on machine learning infrastructure, currently scaling AMD’s ROCm inference and training stack for Instinct and consumer GPUs. With about five years of experience spanning internships and research at Anyscale and Purdue, he has practical expertise in distributed ML, RLHF, offline RL, and GPU-accelerated training. He’s an active open-source contributor to major RL projects—PettingZoo, RLlib, and OpenAI Gym—where he’s added continuous action support, off-policy evaluation features, and multi-agent environment integrations. Rohan’s work blends systems-level engineering (distributed Ray integrations, multi-node fine-tuning) with research-driven RL innovations, and he’s also documented and clarified tooling like Optuna for broader use. Outside work he’s an avid climber, squash player, and pianist, reflecting a balance of discipline and creativity that informs his engineering approach.
5 years of coding experience
2 years of employment as a software developer
Bachelor of Science - BS, Computer Engineering, 3.80, Bachelor of Science - BS, Computer Engineering, 3.80 at Purdue University12th grade, Physics, Chemistry, Math, English, CS, 92%, 12th grade, Physics, Chemistry, Math, English, CS, 92% at Pace Jr.Science College
Github Skills (24)
environmental10
pytorch10
dev-environment10
multi-agent-reinforcement-learning10
python10
rllib10
testing10
gymnasium10
openai-gym10
machine-learning10
reinforcement-learning10
enviroment10
api10
distribute10
environ10
Programming languages (3)
Github contributions (5)
eugenevinitsky/sequential_social_dilemma_games
Sep 2021 - Dec 2021
Repo for reproduction of sequential social dilemmas
Role in this project:
Full-stack Developer
Full-stack DeveloperContributions:57 commits, 4 PRs, 2 comments in 3 months
Contributions summary:Rohan contributed to the development of a PettingZoo wrapper, indicating an effort to integrate the sequential social dilemma games with the multi-agent environment framework. They modified environment code (map\_env.py, cleanup.py, and switch.py), and implemented tests, showing an understanding of environment interactions, and adding functionality to the core game logic. The changes include supporting and testing the newly implemented environment.
sequentialreproductionmachine-learningsocial-dilemmas
ray-project/ray
Jul 2021 - Jan 2023
Ray is an AI compute engine. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads.
Role in this project:ML Engineer
ML EngineerContributions:63 reviews, 15 commits, 47 PRs in 1 year 6 months
Contributions summary:Rohan primarily contributed to the RLlib library within the Ray project, focusing on improvements and fixes for off-policy evaluation (OPE) methods and related components. They addressed issues in existing estimators, such as DirectMethod and DoublyRobust, as well as contributing to new features such as the addition of OPE in evaluation configuration and tests. The user's work also involved modifying and testing the FQETorchModel.
pythonconsistsruntimetensorflowserving
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.