At CVS, I designed a multimodal SNAP Decision Engine combining computer vision, OCR, and LLM reasoning for eligibility decisions, generating approximately $1M in annual business savings. I also optimized search infrastructure, reducing latency from 9 seconds to 150 milliseconds.
I led agentic AI workflows for enterprise search and built ML-based search and recommendation systems using semantic retrieval and ranking. Human-graded evaluations helped reduce hallucinations by 35%, while vector database tuning improved retrieval precision by 28%.
At Skyline Infrastructure, I developed a RAG chatbot that reduced support tickets by 60% while handling more than 500 daily queries. I also built and evaluated predictive models and used A/B testing to assess product changes.
Earlier, I built microservices and a Kafka event pipeline at Cognizant, and developed data and ML pipelines at Florida Fish & Wildlife Conservation. I’m also building Vinge, a video-first dating application using AI to detect inappropriate content.

