At Rest The Case (Lexicon Group), I shipped core AI features for legal document summarization, case research, and timeline generation. I build production agents and multi-step workflows with LangChain and LangGraph.
I designed the RAG layer for PDFs and scanned documents, using chunking, embeddings, FAISS search, and reranking to ground answers in source material. I also grounded multi-turn research in case documents and conversation history.
I built an evaluation practice with DeepEval in CI as a release gate, alongside live checks for grounding, hallucination rate, and latency. I also developed model routing and fallback with LiteLLM, plus prompt caching to reduce latency and token costs.
I built the production FastAPI service and REST endpoints, along with asynchronous document pipelines for S3 ingestion, OCR, and text extraction. My projects include KnowledgeForge, an enterprise AI knowledge platform, and AgentEval, a platform for evaluating LLM agents on quality, cost, and latency.
