M.S. CS @ USC • Full-stack + LLMs/RAG/AI Agents
- Building practical AI systems end-to-end: LangChain/LangGraph, Graph-RAG, FastAPI, Next.js
- Recent: RL-tuned long-context gen (PPO), custom CUDA/Triton speedups, VLM (PaliGemma + SigLIP)
Tech: JS/TS · React/Next.js · Tailwind · Node.js · MongoDB/PostgreSQL/Prisma · PyTorch · FastAPI · Neo4j · FAISS/Qdrant · Docker · GitHub Actions
- Hybrid Graph-RAG for Recruiting — Neo4j + vector search (Qdrant/FAISS), LangChain/LangGraph orchestration, FastAPI backend, Next.js frontend.
- WiM: RL-Enhanced Long-Context Generation — PPO loop for margin drafting, custom CUDA/Triton kernels for speedups, metrics with BERTScore.
- Vision-Language Stack (PaliGemma + SigLIP) — custom tokenizer/processor alignment, image-token handling, KV-cache generation, vocab resize utilities.
- Agentic HR Planner — tool-using agent (search, checklist, Notion API), file-backed memory, JSON-structured outputs.
Skills: JS TS REACT TAILWIND NODEJS MONGO SQL ML CV LLMs RAG AI AGENTS LANGCHAIN LANGGRAPH NEO4J QDRANT/FAISS FASTAPI CUDA TRITON