OnFinance AI

1. 01Built ComplianceOS end-to-end (DB → backend → cloud → LLM pipelines), cutting BFSI compliance effort by 1,500+ hours/year; AI features contributed to $800K+ ARR growth.
2. 02Owned end-to-end design and migration of a custom RBAC engine (V1→V5) — expression DSL compilation, field-level restrictions, and async context resolution for multi-tenant row-level access over regulatory documents.
3. 03Diagnosed and resolved a production OOMKill on a live document parser; root-caused full-document rasterization holding ~4.9 GB RAM, batched processing cut peak memory by 93%, verified and deployed same day.
4. 04Eliminated N+1 query patterns across high-traffic service paths via bulk MongoDB $in queries and iterative BFS subtask traversal — 95–99% reduction in DB round-trips with zero API or schema changes.
5. 05Replaced in-process concurrency limits with message-queue back-pressure for the AI agent service, decoupling HTTP handlers from run executors and eliminating silent job loss on pod restarts.
6. 06Migrated core auth layer from synchronous ODM to async Motor/Beanie across 13 files, resolving event-loop blocking and unawaited-async ordering bugs in audit logs.
7. 07Engineered LangGraph agentic workflows with LiteLLM processing 1B+ tokens/month of SEBI, RBI, and IRDAI regulatory data; built production RAG over financial regulations with versioned embeddings and citation grounding for clause-level retrieval over DRHPs and compliance circulars.
8. 08Contributed to securing $4.2M Pre-Series A from PeakXV Partners through technical demos and executive presentations of ComplianceOS.

Python · FastAPI · LangGraph · LiteLLM · MongoDB · PostgreSQL · AWS · Docker · Kubernetes · RabbitMQ · JWT · OAuth