Technical Lead · M2P Fintech

Document Intelligence Platform

Multi-model document intelligence pipeline trained for 120+ countries, live across 10+, processing 3M+ documents/month at 99%+ extraction accuracy.

What I shipped

• 01

  New document types onboarded in under 2 hours via CNN classifier + LayoutLMv3 fine-tuning

• 02

  Hybrid retrieval on OpenSearch + FAISS with S3 + PostgreSQL storage

• 03

  LLM-augmented extraction via LangChain, LlamaIndex, OpenAI, Bedrock & Gemini

• 04

  FastAPI microservices on Docker / AWS