Work

CVS Health · Scottsdale, AZ

Clinical Data Interoperability Team

• End-to-end development and deployment of an Agentic RAG system which helps clinicians speed up document review process.
• Established an end-to-end ingestion and retrieval pipeline with Mistral OCR preprocessing, semantic chunking, BM25 + dense retrieval, backed by HNSW indexes for low-latency vector search.
• Engineered agent harness: tool definitions, context assembly, memory management, multi-turn conversations and guardrails using LLM-as-a-judge and golden dataset evaluation.
• Implemented an agent observability layer using OpenTelemetry and Grafana.
• Built an NL->Query AI Agent which connects to different data sources and answers busines questions.
• Made the agent self-improving by capturing user feedback on generated queries, and cut cost and latency via complexity-based model routing, query caching, and configurable schema registries.

CVS Health · Scottsdale, AZ

Built and maintained GCP ETL pipelines processing 1M+ weekly transactional records to support upstream ingestion and downstream ML modeling and analytics workflows.

W. P. Carey School of Business, ASU · Tempe, AZ

Answered the question: “What drives innovation in universities?” Merged heterogeneous institutional datasets in pandas, then applied machine learning to isolate which factors predicted innovation output. The work led to a publication in the Journal of Information Systems Education.

With Joseph Cazier and Andres Diaz Lopez.

Dell Technologies · Bengaluru, India

On the Enterprise Infrastructure Planning Tool team. Proposed and built an email classifier using BERT and NLTK to route customer queries automatically, reaching 90% accuracy. It was picked to present at Dell's internal hackathon. Also automated MongoDB extraction, parsing and serialization workflows in C# and Python.