Costa Huang
Member Of Technical Staff at Periodic Labs
Redwood City, California, United States
Summary
🤩
Rockstar🎓
Top SchoolCosta Huang is a machine learning engineer and Member of Technical Staff with a decade of experience building RL and ML systems, currently focused on exploiting physical rewards at Periodic Labs. He has a strong open-source footprint—contributing to high-profile projects like Hugging Face's TRL, OpenAI Gym, and Brax—where he improved RL training utilities, test coverage, and simulation integration. Costa blends research rigor (PhD-level work) with production instincts, shipping PPO and A2C implementations, distributed training improvements, and robust experiment-tracking integrations with W&B. Comfortable across backend systems, CI, and large-scale simulation, he has repeatedly stabilized algorithms and tooling for reproducible benchmarks. Colleagues would note his knack for turning research ideas into portable engineering artifacts that accelerate downstream adoption.
10 years of coding experience
5 years of employment as a software developer
Doctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at Drexel UniversityBachelor of Science - BS, Bachelor of Science - BS at Furman University
Stackoverflow
1,444reputation
463kreached
14answers
2questions
jupyter-notebook
top-5%
Github Skills (42)
unit-testing10
simulation10
pytorch10
pytest10
simulator10
python10
ata10
gymnasium10
simulations10
openai-gym10
machine-learning10
reinforcement-learning10
multiple-gpu10
sagemaker10
huggingface-transformers10
Programming languages (19)
Github contributions (5)
vwxyzjn/cleanrl
Jun 2019 - Jan 2023
High-quality single file implementation of Deep Reinforcement Learning algorithms with research-friendly features (PPO, DQN, C51, DDPG, TD3, SAC, PPG)
Role in this project:
Back-end Developer
Back-end DeveloperContributions:19 releases, 264 reviews, 1302 commits in 3 years 7 months
Contributions summary:Costa implemented core features for the A2C algorithm within the Deep Reinforcement Learning framework. Specifically, they added an A2C implementation with a focus on a CNN-based architecture for Atari environments, and the implementation of a new ReplayBuffer. Furthermore, they formatted existing experiments and contributed to a general codebase used across various RL algorithms.
pythondeep-reinforcement-learninggomokutd3reinforcement
huggingface/trl
Jan 2023 - Aug 2024
Train transformer language models with reinforcement learning.
Role in this project:Full-stack Developer
Full-stack DeveloperContributions:3 releases, 158 reviews, 120 PRs in 1 year 7 months
Contributions summary:Costa contributed to the `trl` repository by implementing and improving several examples and utilities. They worked on adding a Slurm utility for benchmark experiments and an autotag feature with wandb integration. Additionally, they refactored code related to gradient accumulation and made changes to improve stability by changing hyperparameters. They also contributed to the CI and documentation.
nlptransformer-architecturelanguage-modelfinetuninglanguage-modeling
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.