I build LLM-powered applications and the infrastructure behind them — agentic RAG pipelines, AI agent workflows, and ML systems (compilers, accelerators, distributed training) that make large models practical to run.
Python, PyTorch, TensorFlow, LangChain, Hugging Face, FastAPI, Docker
- VerityRAG — Agentic RAG pipeline with a verifier agent: hybrid BM25 + dense retrieval, cross-encoder reranking, cited answer synthesis, and claim-level verification, evaluated on SQuAD v1.1.

