Swagata Sur

I’m an Applied AI and Data Scientist with 3 years of experience at Société Générale, specializing in fraud analytics, automation, and model risk monitoring. I architect end-to-end ML, GenAI, and Agentic AI systems—especially RAG and LLM-powered workflows—along with model monitoring frameworks that turn operational processes into reliable, scalable automation.

At Société Générale, I spearheaded an end-to-end LangGraph-orchestrated document intelligence system using GPT-4.1, improving answer faithfulness and reasoning quality by 35% and reducing hallucinations in AI-generated responses. I also built production-ready sanctions/embargo monitoring (6+ KPIs for drift detection) and an alert similarity scoring pipeline that reduced duplicate AML review effort. For GenAI anomaly classification, I led an end-to-end hybrid rule-based + AI system across 10+ business units, cutting manual categorization effort by 1.5 FTE, and I optimized AML threshold calibration to reduce alert volume by 25% across 16 entities.