Ilya Kostrikov
Member Of Technical Staff (Research Scientist) at OpenAI
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolIlya Kostrikov is a research scientist and Member of Technical Staff at OpenAI with 11 years of experience applying reinforcement learning and generative models to real-world systems, currently focused on RLHF, LLM post-training, tool use, and product-driven safety. His background spans postdoctoral work at UC Berkeley and multiple research roles and internships at Google and Facebook, where he bridged robotics, diffusion models, and large language models. Ilya is a hands-on ML engineer and prolific open-source contributor—his PyTorch implementations of A2C/PPO/ACKTR and A3C are widely used references for reinforcement learning practitioners and demonstrate deep practical knowledge of algorithm design and training pipelines. He has a PhD in Computer Science from NYU and a strong quantitative foundation from earlier degrees in media informatics and applied mathematics. Notably, he combines production-facing deployment experience with core RL research, enabling feature integrations that improve user retention and model performance predictions. Colleagues describe him as someone who moves fluidly between prototyping novel algorithms and shipping robust, well-instrumented systems.
11 years of coding experience
1 year of employment as a software developer
Specialist Applied Mathematics and Computer Science, Specialist Applied Mathematics and Computer Science at Immanuel Kant Baltic federal universityDoctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at New York UniversityMaster of Science (MSc) Media Informatics, Master of Science (MSc) Media Informatics at RWTH Aachen University
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (11)
pytorch10
ppp10
deep-learning10
python10
ata10
reinforcement-learning10
atari260010
machine-learning9
algorithms8
openai-gym7
gymnasium7
Programming languages (6)
Github contributions (5)
ikostrikov/pytorch-a2c-ppo-acktr-gail
Sep 2017 - Nov 2021
PyTorch implementation of Advantage Actor Critic (A2C), Proximal Policy Optimization (PPO), Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation (ACKTR) and Generative Adversarial Imitation Learning (GAIL).
Role in this project:
Full-stack Developer
Full-stack DeveloperContributions:204 commits, 50 PRs, 216 pushes in 4 years 3 months
Contributions summary:Ilya contributed to the PyTorch implementation of various reinforcement learning algorithms, including A2C, PPO, and ACKTR. Their work focused on refactoring the code to incorporate PPO, add visualization features, and integrate KFAC for improved performance. Additionally, the user added support for MuJoCo environments and made changes to optimize the code and improve readability.
policytrust-regionsecond-orderadversarialnatural-gradients
ikostrikov/pytorch-a3c
Feb 2017 - Mar 2019
PyTorch implementation of Asynchronous Advantage Actor Critic (A3C) from "Asynchronous Methods for Deep Reinforcement Learning".
Role in this project:ML Engineer
ML EngineerContributions:38 commits, 11 PRs, 41 pushes in 2 years 1 month
Contributions summary:Ilya primarily contributed to the model architecture and training procedures within the A3C reinforcement learning framework. Their work included modifying the network architecture, debugging and fixing issues related to PyTorch versions, and improving the training process. The user also focused on integrating tools for evaluating the model's performance. The commits show a deep understanding of reinforcement learning concepts, specifically focusing on the implementation and fine-tuning of an A3C model.
pytorcha3cactor-criticasynchronousreinforcement-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.