Sai supriya

About

Built and productionized an AWS-hosted, Kubernetes-orchestrated RAG assistant that enables natural-language Q&A over internal document repositories with grounded answers and citations. Demonstrates strong applied LLM engineering: hallucination mitigation, hybrid retrieval + re-ranking, and rigorous evaluation via benchmarks and A/B testing, plus real-world scaling of compute-heavy inference with dynamic batching and monitoring.

Key Strengths

• Built and deployed production RAG system with citations over large internal document repositories
• Mitigated LLM hallucinations via strict context grounding and citation checks
• Scaled compute-intensive LLM inference using Docker + Kubernetes autoscaling on AWS
• Applied inference performance techniques (dynamic batching) with production monitoring
• Experiment-driven model selection using benchmarks and A/B testing (accuracy, hallucination rate, latency)
• Improved retrieval quality by pivoting from dense-only to hybrid (vector + BM25) for technical documents
• Effective collaboration with product managers using demos/prototypes and translating technical concepts into business terms

Experience

Education

Skills

Languages