Jayavardhan kolla

I build and deploy production LLM applications and agentic RAG workflows, along with backend AI services. I focus on reliable delivery—tool calling, structured outputs, streaming, semantic caching, guardrails, and LLM evaluation—using Python, FastAPI, LangGraph, and cloud stacks like AWS Bedrock with vector search (Qdrant) and Redis.

In my current work, I’ve built LangGraph-based multi-step agents that use tool calling across documents, Salesforce, and SQL, including query validation, retries, and error recovery. I also lead end-to-end ingestion through chunking, embeddings, metadata-aware retrieval, and vector-search pipelines, and I’ve delivered measurable outcomes such as reducing clinician documentation time by 70% and improving note quality (92% accuracy), while a large-scale NLP validation tool achieved 98% rule adherence and cut review time by 30% through close collaboration with radiologists.