Aleksei Petrenko
Senior Research Scientist at Apple
Los Angeles, California, United States
Summary
🤩
Rockstar🎓
Top SchoolAleksei Petrenko is a Senior Research Scientist at Apple with 11 years of experience at the intersection of deep reinforcement learning, simulation, and robotics, and a PhD from USC. He has a strong track record of building high-throughput RL systems and simulators—co-authoring influential open-source projects like Sample Factory and Megaverse that enable massive single-node training and million-FPS multi-agent simulation. His work spans from dexterous manipulation and quadrotor swarms to LLMs and ML systems, combining low-level systems engineering (C++/Vulkan/OpenGL) with state-of-the-art ML research (PyTorch, population-based training, self-play). Known for practical, production-ready research, he routinely bridges academic publication and industrial deployment, with internships at NVIDIA and Intel Labs that led to multiple conference submissions. Based in Los Angeles, he brings deep domain expertise in sim-to-real transfer and GPU-accelerated experimentation, often optimizing tensor and rendering pipelines that are easy to overlook but crucial for scaling RL.
11 years of coding experience
10 years of employment as a software developer
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at University of Southern CaliforniaBachelor's degree, Computer Science, 4.96/5.00, Bachelor's degree, Computer Science, 4.96/5.00 at Nizhniy Novgorod State Technical University named after R.Y. Alekseev (NSTU)
English, Russian
Stackoverflow
6,968reputation
683kreached
66answers
103questions
docker
top-5%
emulation
top-5%
windows
top-5%
Github Skills (18)
pytorch10
operation10
python10
tensorrt10
machine-learning10
reinforcement-learning10
tensorflow10
model-optimization10
tensor10
windows9
docker9
gpu-programming9
emulation9
virtualization6
haxm6
Programming languages (6)
Github contributions (5)
alex-petrenko/sample-factory
Jun 2019 - Jan 2023
High throughput synchronous and asynchronous reinforcement learning
Role in this project:
ML Engineer
ML EngineerContributions:1 release, 238 reviews, 1194 commits in 3 years 7 months
Contributions summary:Aleksei's contributions focused on enhancing the DMLab environment, specifically for multi-GPU rendering, optimizing observation preprocessing, and integrating a new KL-divergence-based exploration loss. They refactored the code to improve tensor operations, added support for a different set of actions, and fixed issues related to invalid actions. The user also worked on ensuring correct training, testing, and general functionality in the multi-agent environment.
asynchronousdeep-learningreinforcement-learninghigh-throughputthroughput
alex-petrenko/faster-fifo
Mar 2020 - Jul 2022
Faster alternative to Python's multiprocessing.Queue (IPC FIFO queue)
Contributions:4 releases, 14 reviews, 57 commits in 2 years 4 months
pythonfifo-queuemultiprocessingfasterfifo
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.