Hi, I'm Shubham.

Most of my week goes to one question: given a fleet, a pile of orders and a long list of business rules, what is the best plan? Our Vehicle Routing Problem solver uses Iterated Local Search and runs on Kubernetes in seven regions. I've turned more than fifteen enterprise requirements into solver constraints, among them EV charging, multi-trip reloads, time windows, driver shifts and LTL/FTL planning, and I take each one from scoping through UAT sign-off.

Lately I've been building AI agents on top of that work. Pilot Core automates a dispatcher's day and cut route planning from six hours to twenty minutes. I also brought Claude Code, Cursor and MCP-connected agents into the routing team's daily work for building, code review and root-cause analysis.

Before routing I worked across the stack. I built FarEye Track, a real-time shipment visibility platform on Kafka and Elasticsearch that grew to 20+ enterprise clients, and an ML model that predicts delivery dates with 95% accuracy. Earlier, at Wynk Music, I grew organic traffic from 7K to 700K+ through server-side rendering and SEO work.

I mentor 13 engineers through design reviews, and six of them have been promoted to senior roles. That part of the job matters to me as much as the systems.