At Databricks, I built RAG workflows for internal knowledge search that ground LLM answers in source documents, reducing unsupported answers by 30% on an internal test set. I also tuned retrieval settings across Pinecone and PGVector, improving retrieval relevance from 0.70 to 0.85.
I developed Spring Boot and Spring AI services for LLM features, with response validation and fallback handling. I containerized services with Docker and deployed them to AWS through Jenkins and GitHub Actions, cutting release time from 40 minutes to under 15 minutes.
At Accenture, I integrated machine-learning inference into Java and Spring Boot microservices used by four enterprise applications. Earlier at NVIDIA, I converted AI research prototypes into application components and internal tools used by engineers and researchers, and built APIs that delivered model outputs to internal applications.
