Senior Data Scientist | GenAI, Data & Analytics
NeoGenCode Technologies Pvt Ltd · Feb 2025 – Present
- PinakaAi - KYC Document Intelligence Microservice
- Care Health Insurance
- AI-powered KYC microservice that masks PII on Aadhaar, classifies and extracts fields from 5+ Indian ID types using YOLO + PaddleOCR + AWS Bedrock, with JWT-secured APIs and PostgreSQL audit trails.
- -> Tech Stack :
- Backend & API -: Python, FastAPI, PostgreSQL, Alembic, PyJWT, Cryptography(RSA/AES), AWS, Docker).
- AI/ML -: YOLO, PaddleOCR, OpenCV, PyTorch, Numpy, AWS Bedrock (Claude 3 Haiku).
- Cloud & MLOps -: AWS(ECR, Bedrock, S3 via DocService), boto3.
- Integration & Security-: JWT authentication, RSA+AES encrypted API payloads, client credential validation, CORS.
- Supported docs-: Aadhaar, PAN, Passport, Driving License, Voter ID
- 🔹Led end-to-end delivery of PinakaAI for Care Health Insurance — a production KYC microservice combining computer vision, OCR, and AWS Bedrock LLM for document masking, classification, and field extraction.
- 🔹Architected a dual-service system (Starlette backend + FastAPI AI service) deployed on AWS ECR/Docker, integrated with DocService (S3), AnZ auth, and PostgreSQL for stage-wise processing audit trails.
- 🔹Built and mentored a team of 5 on the AI pipeline (YOLO, PaddleOCR, Bedrock); owned deployment across Dev/UAT/Prod environments.
- 🔹Processed 2M+ KYC documents and masked ~1M Aadhaar cards, compliant with Indian UIDAI PII masking requirements.
- 🔹Improved classification accuracy for edge cases (e.g.
- Aadhaar front vs back) using an LLM-priority architecture over pure computer vision.




