At Pibit.ai, I architected Prism, an event-driven reporting platform that serves 20+ clients’ report formats from one codebase. I also built a workflow engine that lets customers compose document-processing pipelines from reusable steps.
I built a RAG chat assistant for underwriters to query large, multi-format submission packets, along with its document indexing pipeline. I also shipped agent-driven Excel templatization that generates reports from client templates.
At Digitap.ai, I led the 10-person extraction and fraud-detection team behind the lender API. I integrated the Account Aggregator framework and built tamper detection for loan statements, cutting false positives from 6% to 0.8%.
I also engineered and parallelised the core PDF extraction engine across bank formats, reducing processing time for 60-page statements from 30 to 5 minutes. My work spans Python platforms for enterprise SaaS in insurance and fintech.
