At Node2.io, I design and ship AI-powered workflows for BuildingSync, including an LLM Copilot and a complaint-triage engine that ranks issues by severity. I added a human review step before automated actions are committed.
I built the platform's multimodal data layer for elevator telemetry, anomaly detection, occupancy trends, and document extraction. I also own its production backend, including async FastAPI services and multi-tenant PostgreSQL with Row-Level Security, deployed through CI/CD on AWS.
I built and published Scout, an open-source Python security scanner on PyPI, with an MCP server for agent-driven scan, fix, and re-verify workflows. For an inbound phone booking agent, I implemented address read-back confirmation and webhook calls to calendar and maps services, and wrote a reference implementation with 962 passing tests.
My research includes landslide segmentation using Sentinel-2 imagery at the Defence Geoinformatics Research Establishment and EEG signal processing at Indian Institute of Technology (IIT). I’ve also co-authored two IEEE publications.
