At Insurance Samadhan, I re-architected the KYP engine around event-driven pipelines and concurrency control, cutting end-to-end latency by about 50% while scaling to 20K+ requests a day. I also built a Gemini-based LLM and RAG pipeline for insurance documents that reached 94% field-level extraction accuracy.
I shipped production AI agents for WhatsApp document intake and expert review, and engineered Kafka microservices and FastAPI inference APIs running at 99.9% uptime. I also reduced GCP costs and document-pipeline storage through infrastructure and data-processing changes.
On Sley, an early-stage consumer social-commerce product, I'm the sole backend and AI engineer building multi-modal product search, recommendations, and creator attribution. I also built Companion-AI, a long-term memory and consistency engine with an evaluation harness that improved results from 88% to 100% across its probes.
