About me
I'm Shivam Batra, a software engineer with 10 years of experience building backend systems at scale, including at Delhivery and Ula. I now work on reliable AI agents, with a deep focus on evals, inference performance, and LLM optimization through KV caching, benchmarking, and careful system design. My background in logistics, payments, inventory, and warehouse platforms helps me build AI systems that hold up in real workflows.
Loading writings…
Loading project updates…
Experience
At Delhivery and Ula, I built backend platforms where throughput, correctness, and reliability had direct consequences for teams in the field. That engineering foundation now informs how I design and evaluate AI agents.
Delhivery
I worked on first-mile platforms handling 5M+ shipments a day, a payouts platform for 100K+ users, and AI-driven dispatch. These systems connected routing, exception handling, operational decisions, and financial flows at scale.
Ula
I built procurement and warehouse systems, improved inventory accuracy through inter-warehouse transfers, and developed order throttling infrastructure. The work required dependable data flows and a clear understanding of how stock and orders move.
Delhivery
Team
Ula
Field notes
BuildingToday I apply that experience to reliable AI agents: defining useful evals, measuring inference performance, benchmarking model behavior, and improving LLM serving with techniques such as KV caching. I care about what an agent does under real operating conditions, not just in a demo.
Projects
Recent commits
GitHub ↗Loading recent work from GitHub…
Contact