Sumit Chintanwar

I've built distributed systems ranging from a SQL query engine and workflow platform to a multi-modal LLM inference platform and real-time market data infrastructure.

In Inference Mesh, I built a Go and Python platform with OpenAI-compatible streaming APIs, Kubernetes scaling, JWT authentication, rate limiting, and health-aware worker routing. I reduced gateway-to-worker connection setup latency by 88% and validated 454 ms P95 latency with 291 sustained requests per second under 100 concurrent users.

For Velocity Feed, I designed a Go, Redis Pub/Sub, and WebSocket platform that delivered 0.18 ms P50 and 0.67 ms P99 latency across 1,000 concurrent clients while sustaining roughly 5,000 publish messages per second with zero dropped messages.

I've also contributed an upstream C/C++ optimization to OSGeo GRASS GIS that reduced r.watershed runtime by 75% on million-cell raster datasets. I enjoy solving systems problems across concurrency, networking, performance, observability, and production reliability.