- π Ph.D. Student in Electrical Engineering (Moody Graduate Fellow) at Southern Methodist University
- πΌ Former Research Scientist Intern at AT&T (Chief Data Office)
- π¬ Researching LLM Pre/Post-Training & Alignment, Low-Precision Inference Systems (NVFP4/CUDA), and Agentic Systems
- π§ Featured Work: Predicting ID with LLaMA
- π« How to reach me: jerryma@smu.edu
- π Languages: Python, C/C++, CUDA C++, Rust, TypeScript, Bash, SQL
- π§ ML & Systems: PyTorch, Hugging Face (Transformers, PEFT, TRL), vLLM, TensorRT-LLM, DeepSpeed, Ollama
- π€ Agentic AI & Retrieval: Model Context Protocol (MCP), LangChain, GraphRAG, Qdrant, Neo4j
- π Tools & Infrastructure: Linux, Slurm HPC, Docker, Git, AWS (EC2), GCP
- kug_eval - Knowing-Using Gap (KUG) diagnostic toolkit for SFT generalization & memorization
- graph_rag - Graph-vector database for tree/graph-of-thought knowledge representation and agentic reasoning
- PyTAO - A toolkit to assist ONIOM calculations in Python
- Automative Debug Assistant - LLM-powered debugging assistant for custom software projects
β Thanks for stopping by! Feel free to check out my repositories or connect with me on LinkedIn.



