At Emplay Analytics, I lead the production Obligation Management LLM microservice, which processes contract and purchase-order streams to determine claims and claim amounts. Its evaluation pipeline reached 96% accuracy, and concurrent processing handles 1,600 PO rows in about 20 minutes with zero data loss.
I also built multi-retriever RAG architectures that improved document retrieval precision by 40%. At Emplay Analytics, I developed metadata-driven ingestion pipelines and low-code engine components that cut enterprise workflow development cycles by 50%.
In my projects, I built PromptPulse to monitor LLM quality, latency, and cost, and used semantic caching to reduce P95 latency by 45%. I’ve also built an agentic restaurant reservation system, a chest X-ray pneumonia detector, and a LangGraph analytics agent.
