Skip to content
View yixinhuang48's full-sized avatar
๐Ÿ 
Working from home
๐Ÿ 
Working from home
5 followers · 32 following

Achievements

Achievement: Pull SharkAchievement: Quickdraw

Achievements

Achievement: Pull SharkAchievement: Quickdraw

Block or report yixinhuang48

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this userโ€™s behavior. Learn more about reporting abuse.

Report abuse
yixinhuang48/README.md

Hi, I'm Yixin ๐Ÿ‘‹

Typing SVG

I work on LLM systems, evaluation, and GPU-accelerated ML infrastructure.
RA @ UCSD Hao AI Lab | M.S. Computer Science student
๐Ÿ“ San Diego, CA | Focused on large-scale training, inference, and agent evaluation.

Personal Website Lab Website GitHub Zhihu

Visitor Count

๐Ÿ† GitHub Achievements: Pull Shark Quickdraw

"A journey of a thousand miles begins with a single step." โ€” Confucius

๐Ÿ’ฌ Random Dev Joke

Jokes Card

๐Ÿ”ฌ Research & Systems Interests

  • LLM evaluation & benchmarks (agents, games, scientific reasoning)
  • Large-scale training & inference systems (FSDP, vLLM, Ray, Slurm)
  • GPU efficiency, memory systems, and model parallelism
  • Reinforcement learning for agents (GRPO, NeMo-Gym)

Tech Stack:

Python PyTorch CUDA Docker Ray Slurm Git Linux Jupyter vLLM SGLang NeMo RL Areal

๐Ÿ›  Selected Projects

๐ŸŽฎ GamingAgent โญ 843
LLM/VLM gaming agents and model evaluation through games
โ†’ long-horizon reasoning, memory & perception harnesses
(Doom, Sokoban, Tetris, Pokรฉmon Red)

๐Ÿ”ฌ VideoScience โญ 5
Benchmark for scientific correctness in text-to-video models
โ†’ physics & chemistry concepts, VLM-as-Judge scoring
(CVPR submission)

๐Ÿค– NVIDIA NeMo Gym โญ 603
Build RL environments for LLM training
โ†’ scalable RL training, reward profiling, GRPO
Integrating Sokoban & Tetris

๐ŸŒ lmenv
LLM environment framework for interactive evaluation
โ†’ standardized interfaces for game-based agent testing

GitHub Activity Graph

๐Ÿง  Current Focus

  • ๐Ÿ”„ Scaling agent evaluation with interactive environments
  • โšก Training & serving efficiency on multi-GPUs
  • ๐ŸŽฏ Reward modeling and RL for LLM agents

๐Ÿ“š Currently Learning

  • Advanced distributed training techniques (FSDP, DeepSpeed)
  • GPU memory optimization and profiling
  • Large-scale RL systems architecture

๐Ÿ”— Connect with Me

๐Ÿ“ˆ Profile Summary

Profile Summary Card

Pinned Loading

  1. hao-ai-lab/VideoScience hao-ai-lab/VideoScience Public

    Python 5 1

  2. lmgame-org/GamingAgent lmgame-org/GamingAgent Public

    [ICLR 2026] LLM/VLM gaming agents and model evaluation through games.

    Python 849 90

  3. lmgame-org/lmenv lmgame-org/lmenv Public

    Python 1

  4. NVIDIA-NeMo/Gym NVIDIA-NeMo/Gym Public

    Build RL environments for LLM training

    Python 621 53

  5. yixinhuang48.github.io yixinhuang48.github.io Public

    CSS 1