At Concentrix, I built risk-scoring models for agent fraud and work avoidance, combining statistical thresholds with Isolation Forest to flag risk across five tiers. An interactive Power BI dashboard gave business users visibility into high-risk agents and avoidance patterns.
For Kotak Life Insurance, I built persistency, revival, and surrender-probability models to support retention strategies. The work contributed to a 9% improvement in policy revival rates and a 1% increase in premium collection.
I developed a contract document intelligence pipeline using Azure Logic Apps, Databricks, NLP, and a customized open-source LLM. Its structured insights reduced manual contract-review effort by an estimated 50–70%, and I explored agentic workflows for more autonomous review.
I also built a utility-bill extraction pipeline that captured more than 117 structured fields from PDFs at an estimated 94% field accuracy. I shipped it as a Streamlit application, containerized with Docker, for uploads and structured Excel exports.
