saurav kumar

I’m a Senior Software Engineer at Vitrana (A Parexel company), where I design and build production-grade Python services exposed through FastAPI REST APIs and microservices. I help ensure reliability and scalability while supporting high-throughput production workloads (10,000+ requests/month).

I architect Retrieval-Augmented Generation (RAG) pipelines over clinical and regulatory documents using LangChain and vector databases with semantic embeddings and hybrid search. By integrating evaluation and regression suites that measure groundness, hallucination rate, and prompt/model performance, I improve release confidence and retrieval relevance.

I build AI-powered development workflows using Claude Code and multi-agent orchestration, leveraging MCP, custom agent skills, hooks, and tooling standards to accelerate delivery. I also integrate Model Context Protocol (MCP) servers to standardize tool and data access for LLM agents across internal systems, reducing effort by 50%.

On the model and automation side, I fine-tune open-source SLMs with LoRA/QLoRA for domain-specific clinical data to improve extraction and NER accuracy while reducing inference cost. I’ve also developed an agentic unit testing framework and document-to-data extraction services that significantly reduce manual work (60% product coverage automated; 40% unit testing effort reduction; 80% reduction in manual patient case processing).