
Software Development Engineer II | Applied AI
Aguna Solutions · Jan 2026 – Present
- Architected IntelliDMS across 5 enterprise workflows (procurement, contracts, CAPEX, MRF, PE/PO) using async FastAPI, PostgreSQL, Docker, LDAP, and RAG-ready search; centralized document metadata and role-based routing to reduce average document lookup and approval-handoff time by 60% in pilot workflows.
- Implemented an 8-stage unstructured-data pipeline for OCR, chunking, metadata, embeddings, vector indexing, retrieval, and re-ranking; on a 10K-document benchmark, async batching improved ingestion/indexing throughput by 2.0x with 98%+ successful document processing while preserving source lineage.
- Embedded semantic retrieval, role-aware context, PII/output validation, and audit trails across 10+ RBAC roles; hybrid retrieval plus cross-encoder re-ranking reduced unsupported LLM responses by 35% on an internal evaluation set while keeping responses restricted to authorized source documents.
- Built predictive-maintenance analytics and a CNN-based real-time face-recognition application for NIC India, achieving 95%+ recognition accuracy and integrating low-latency inference into a real-time application workflow.


