Huy Le — AI Researcher

Rochester, NY Email: legiahuy050104@gmail.com | GitHub: github.com/hl9082 I'm an AI Research Assistant who optimizes model training pipelines on GPU hardware and publishes peer-reviewed NLP research - as an undergraduate. Most entry-level researchers assist; I co-shipped a publication at EACL 2026, optimized Gemma-270M training by 60% on NVIDIA A100s, and built the experiment tracking infrastructure from scratch. Previously, as an AI Research Assistant at Rochester Institute of Technology's Homan Lab, I increased Gemma-270M model GLEU scores from 0.03 to 0.80 on 50K Zarma examples while cutting training time by 60% and memory usage by 50% using CUDA optimization on NVIDIA A100 GPUs. CAREER HIGHLIGHTS: >> Increased Gemma-270M model GLEU score from 0.03 to 0.80 on 50K Zarma Hugging Face examples by fine-tuning with PyTorch and mixed precision training on NVIDIA A100 GPU >> Reduced model training time by 60% and memory usage by 50% by implementing CUDA optimization, TF32 precision, and PyTorch no-gradient inference >> Led a team of 4 to build 13 MCP tools in 2 days with FastMCP, reducing debugging time by 35-50% on 200 MB GitHub repositories >> Achieved accurate ASL sign prediction across 29 classes by training a deep learning model for 50 epochs with PyTorch on Google Colab Tools: Python, PyTorch, Hugging Face, Transformers, NVIDIA CUDA, scikit-learn, pandas, NumPy, React, FastAPI, Spring Boot, FastMCP, IBM Watson, wandb, Git/GitHub, Google Colab I am an AI Research Assistant who can optimize model training pipelines on GPU hardware AND publish peer-reviewed NLP research - as an undergraduate. I thrive in teams that value technical depth, innovation, and genuine AI research. Email: legiahuy050104@gmail.com GitHub: github.com/hl9082 When I'm not training models, I'm cheering for my favorite wrestling superstars or playing soccer.

Stackforce AI infers this person is a skilled AI Researcher with strong expertise in NLP and model optimization.

Location: Rochester, New York, United States

Experience: 0 mo