M.S. in Electrical and Computer Engineering @ UC San Diego
LLM4Code · AI4SE · Code Agents · Agent Systems · Machine Learning
I work on LLM4Code, AI for Software Engineering, code agents, agentic systems, and machine learning. My current direction is building retrieval-augmented and agentic workflows that can reason over code, repair programs, support verification, and turn research prototypes into reproducible engineering systems.
I am especially interested in the intersection of:
| Area | What I Care About |
|---|---|
| LLM4Code | code reasoning, program repair, evaluation, repository-level context |
| AI4SE | automation for testing, debugging, verification, and developer workflows |
| Code Agents | tool-using agents, iterative repair, long-context planning, execution feedback |
| Agent Systems | retrieval, orchestration, memory, workflow reliability, human-in-the-loop systems |
| Machine Learning | deep learning, representation learning, data-centric evaluation, applied ML |
- ISQED 2026: VeriRAG: A Retrieval-Augmented Framework for Automated RTL Testability Repair accepted.
- ICML 2025: paper accepted.
- IJCAI 2025: paper accepted.
- ML4H Findings 2025: paper accepted.
- TRITB 2025: paper accepted.
- Started the M.S. program in Electrical and Computer Engineering at UC San Diego.
- Built and open-sourced structured course and project archives to make graduate-level learning materials easier to reuse.
| Project | Direction | Why It Matters |
|---|---|---|
| VeriRAG | LLM4Code, AI4SE, RAG, RTL repair | Retrieval-augmented framework for automated RTL testability repair and hardware-oriented code reasoning. |
| ucsd-course-materials | ML, NLP, systems coursework | Organized UCSD graduate course archive with assignments, reports, slides, and project material. |
| AWSG-FITE-Web | research lab web engineering | Frontend and content infrastructure for presenting research, people, and lab activity. |
| IoT-RAG-ABA | RAG, applied AI | Domain-specific retrieval workflow for applied AI systems. |
- Building agentic code-repair systems that combine retrieval, execution feedback, and structured reasoning.
- Applying RAG to verification and repair, especially for RTL and hardware design workflows.
- Designing AI4SE evaluation pipelines where benchmarks, failure cases, and metrics are explicit.
- Turning research code into clean public repositories with usable READMEs, reproducible structure, and clear assumptions.
Research / ML: PyTorch, transformers, retrieval, RAG, evaluation, experiment design
AI4SE / Agents: code repair, tool use, execution feedback, benchmark construction, repository analysis
Engineering: Python, JavaScript, React, Git, reproducible project structure, documentation
Domains: LLM4Code, AI4SE, code agents, RTL repair, applied ML, information retrieval
- Make the problem setting explicit before optimizing the model.
- Treat evaluation as a first-class artifact, not an afterthought.
- Use retrieval and tools when the task depends on external context.
- Prefer systems that can be inspected, reproduced, and improved by other researchers.
- Publish clean references so future students and builders can move faster.
I am gradually organizing my research, coursework, and engineering projects into public repositories around LLM4Code, AI4SE, code agents, RAG, and machine learning. The goal is to make each repository useful as a concrete reference: clear structure, visible assumptions, runnable code when possible, and enough context for future students to learn from it responsibly.