Sumeet Batra
Research Scientist at Ai2
Los Angeles, California, United States
Summary
👤
Senior🎓
Top SchoolSumeet Batra is a research scientist specializing in machine learning for robotics, with a PhD from USC and a decade of experience spanning RL, representation learning, and generative methods for drones, autonomous vehicles, and manipulation. He joins Allen Institute for AI to work on generalized embodied intelligence after a research trajectory that includes high-throughput RL systems and diffusion-based scenario generation at NVIDIA and GAN experimentation at NIST. His contributions to the popular sample-factory RL codebase—adding custom encoders, spectral normalization, and multi-agent support—reflect both systems-level engineering and algorithmic insight. Comfortable moving between simulation-scale engineering and foundational research, he builds scalable pipelines for large-scale experience collection as well as novel models for diverse trajectory generation. Colleagues describe him as intellectually curious about the nature of intelligence and pragmatic about deploying research in realistic simulators.
10 years of coding experience
2 years of employment as a software developer
Doctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at University of Southern CaliforniaBachelor of Science (B.S.) Computer Science, Bachelor of Science (B.S.) Computer Science at University of Colorado Boulder
English, Japanese
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (7)
neural-network10
pytorch10
machine-learning10
python10
reinforcement-learning10
encoder10
deep-learning9
Programming languages (4)
Github contributions (5)
alex-petrenko/sample-factory
Aug 2020 - Feb 2021
High throughput synchronous and asynchronous reinforcement learning
Role in this project:
ML Engineer
ML EngineerContributions:9 reviews, 21 commits, 13 PRs in 5 months
Contributions summary:Sumeet made significant contributions to the reinforcement learning project by implementing and refactoring components related to multi-agent environments and neural network encoders. They developed custom encoders, including a mean embedding encoder, to improve observation space handling and model performance. Their work included refactoring the forward method, adding support for spectral normalization to stabilize training, and extending observation spaces with goal position information.
asynchronousdeep-learningreinforcement-learninghigh-throughputthroughput
SumeetBatra/SumeetBatra.github.io
Sep 2021 - Sep 2021
Contributions:33 commits, 11 pushes, 1 branch in 1 day
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.