Botao Chen
Member Of Technical Staff at Anthropic
Menlo Park, California, United States
Summary
🤩
Rockstar🎓
Top SchoolBotao Chen is a Member of Technical Staff with eight years of experience building large-scale ML and training infrastructure, currently helping make Claude more aligned and safe at Anthropic. Previously at Meta he engineered distributed LLM training systems (Llama3/4) and led recommendation and ranking initiatives for Instagram Reels that materially improved engagement and cohort optimization. He contributes to prominent open-source projects in the PyTorch ecosystem—optimizing training speed, dataset builders, and LoRA SFT workflows in torchtune and llama-stack—demonstrating a mix of production-grade infra and ML research tooling. With an MS from NYU and a BS from Fudan, he combines rigorous academic training with hands-on systems work across research and consumer-scale products. An understated strength is his focus on code quality and measurable validation—unit tests and metrics are a recurring theme in his contributions.
9 years of coding experience
7 years of employment as a software developer
Master of Science - MS, Electrical and Computer Engineering, Master of Science - MS, Electrical and Computer Engineering at New York UniversityBachelor of Science - BS, Electrical and Computer Engineering, Bachelor of Science - BS, Electrical and Computer Engineering at Fudan University
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (13)
lora10
pytorch10
machine-learning10
python10
model-checking9
checkpointing9
apidoc9
checkpoint9
deep-learning9
api9
unit-testing9
natural-language-processing7
nlp7
Programming languages (1)
Github contributions (5)
ogx-ai/ogx
Nov 2024 - Mar 2025
Open GenAI Stack
Role in this project:
ML Engineer
ML EngineerContributions:41 reviews, 62 PRs, 126 pushes in 4 months
Contributions summary:Botao's commits center around integrating torchtune with the llama-stack project for post-training tasks, specifically focusing on single-device LoRA SFT. Their contributions involve refining post-training APIs, implementing a supervised fine-tuning skeleton, and creating a basic LoRA training recipe leveraging torchtune components. The work also includes integrating job management APIs and creating and validating metrics during training.
genai
meta-pytorch/torchtune
Mar 2024 - Dec 2024
PyTorch native post-training library
Role in this project:ML Engineer
ML EngineerContributions:19 reviews, 20 PRs, 27 pushes in 8 months
Contributions summary:Botao primarily focused on optimizing the training process within the PyTorch-based library. Their commits include refactoring code to improve training speed, specifically by adjusting the placement of loss calculations. Furthermore, the user added and modified dataset templates, enhancing the library's ability to handle various instruction datasets and building other flagship instruction dataset builders. They also contributed to unit tests, demonstrating a focus on code quality and validation of the implemented changes.
post-trainingpytorch
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.