Three companies so far. A startup I co-founded, an enterprise blockchain platform where I grew from intern to engineer, and the internship where it started.
We build a knowledge graph over a company's databases, so anyone can ask a question in plain English and get correct SQL back. I lead architecture and engineering end to end.
Knowledge graph pipeline
Reads a database and turns it into meaning in six steps: introspect, profile, cluster, classify, synthesize, graph.
Text-to-SQL agent
Routes each question by intent, writes several SQL candidates, lets them vote, then runs the winner in a sandbox on Postgres, MySQL, SQLite or Snowflake.
Data layer
FalkorDB for the graph, Qdrant for vectors, Redis for queues and streaming, PostgreSQL for the rest. PII detection with Presidio.
SaaS control plane
A license server in FastAPI and React, with licenses signed through GCP Cloud KMS, plus a benchmark harness for the agent.
I joined as an intern and stayed two and a half years, working across the Go backend, the Python data steps and the React apps enterprises used every day.
Control Tower
Designed the event-driven Go service that moves every document from intake, through a queue, Python preprocessing, a NATS cache and RAFT consensus, to the block write. It replaced a much slower Node.js service.
Param Wallet
Led the move from class components to functional React with one state model on Redux and Context. The app now loads several times faster.
Param AI Studio
Built the product UI: agent operators, MCP connectors, chat over databases with RAG, role-based access and Razorpay billing.
Build and deploy
Moved the frontend from Webpack to RsBuild. Deploys finish in under a minute, and rollbacks are near instant.
Internal ticketing
Designed the ticketing platform in React, Node.js and MySQL, with RBAC, a Kafka notification pipeline and analytics dashboards.
ENN dashboard
As an intern, built the user management and activity dashboard and tuned the MongoDB and DynamoDB queries behind it.