junior software Developer
- Built a GenAI + VLM + OCR doc-extraction pipeline (Python, Docker) converting PDFs/Word/scans/handwriting to schema-validated JSON/CSV/XML; ~90% field-level F1, p95 ≤30s/doc.
- Productionized with event-driven ingest and async workers; delivered ~2,000 docs/hr throughput, 99.9% availability, and 91% auto-recovery of transient failures.
- Added confidence-based routing, human-in-the-loop review, PII redaction, and encryption (AES-256/TLS); cut review time –42% with full audit trails.