Yixin Huang

Hi, I'm

Yixin Huang

|

Research Assistant at UCSD Hao AI Lab
M.S. Computer Science Student • San Diego, CA

View Projects 📄 Resume
research.py 
class Researcher:
    def __init__(self):
        self.focus = [
            "LLM Systems",
            "Agent Evaluation",
            "GPU Infrastructure"
        ]
        self.tools = [
            "vLLM", "SGLang",
            "NeMo RL", "Ray"
        ]
    
    def build(self):
        return "🚀 Innovation"
Scroll

About Me

"A journey of a thousand miles begins with a single step." — Confucius

I work on LLM systems, evaluation, and GPU-accelerated ML infrastructure. Currently an M.S. Computer Science student at UC San Diego (GPA: 4.0), previously B.A. in Computer Science & Applied Mathematics from UC Berkeley (GPA: 3.86).

Research 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)

Current Focus

  • 🔄 Scaling agent evaluation with interactive environments
  • ⚡ Training & serving efficiency on multi-GPUs
  • 🎯 Reward modeling and RL for LLM agents

Tech Stack

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

Featured Projects

🎮

GamingAgent

⭐ 843

LLM/VLM gaming agents and model evaluation through games. Evaluates long-horizon reasoning, memory & perception in Doom, Sokoban, Tetris, and Pokémon Red.

Python LLM Evaluation
View Project →
🔬

VideoScience

CVPR

Benchmark for scientific correctness in text-to-video models. Evaluates physics & chemistry concepts using VLM-as-Judge scoring.

Python Benchmark Video
View Project →
🤖

NVIDIA NeMo Gym

⭐ 603

Build RL environments for LLM training. Integrating Sokoban & Tetris for scalable RL training, reward profiling, and GRPO.

Python RL NVIDIA
View Project →
🌐

lmenv

LLM environment framework for interactive evaluation. Standardized interfaces for game-based agent testing.

Python Framework
View Project →

Get in Touch

Feel free to reach out for collaborations, discussions, or just to say hi!

Email GitHub Zhihu Resume

📧 Recruiters: Feel free to reach out at yixinhuang48@gmail.com

💬 Open an issue or discussion on any of my repositories!