At IBM, I build production ML and GenAI systems for supply-chain decisioning, enterprise search, and regulated sustainability reporting.
I led the architecture and deployment of Enterprise DocSearch across 42,000+ documentation pages and 200,000+ retrieval units, combining lexical and semantic search with cross-encoder reranking. I also built MCP-based and task-specific agent frameworks that coordinate AI skills across structured workflows.
I developed document extraction for 120+ lease fields, including OCR for scanned documents and source-linked review workflows. For Scope 3 reporting, I helped build a platform processing approximately 60,000 transactions per client while maintaining regional data residency, traceability, and validation controls.
Earlier at IBM, I deployed an XGBoost model for SKU cancellation probabilities and built node-performance products using predicted order and inventory inflows. My work spans evaluation, temporal validation, probability calibration, deployment, and production monitoring.
