Building distributed UI systems, AI architectures, and platform infrastructure that powers 100M+ sessions/month.
Seven years at Paytm — India's largest payments platform — is not a lack of ambition. It's what happens when you keep being handed harder problems. First building UI primitives, then owning the rendering pipeline, then setting the frontend architecture standards that many product teams now ship on top of.
The shift to Tech Lead was the moment the job became about leverage: the API contracts that gave backend teams a stable surface to build against, the CI/CD performance gates that caught regressions before review, the design system that eliminated duplicated UI effort across every vertical. The goal is always the same — make other teams faster without them knowing why.
That same instinct now applies to AI systems. Not wrappers around model calls, but production-grade orchestration — agentic workflows, RAG pipelines, MCP gateways — built with the same obsession for observability, failure tolerance, and team ergonomics as any platform layer.
"The best platform work is invisible — other teams move faster and don't know why. That's the goal whether it's a shared component system or an LLM routing layer."
Production systems serving hundreds of millions at Paytm — and personal experiments on GitHub that show where Abhay is headed next.
Built the systems behind India's most-used payment platform. Now building the systems behind AI.
A focused set of battle-tested technologies — chosen for reliability, performance, and scale.
Outcomes measured in milliseconds saved, systems shipped, and engineers grown.
Cut page load by 40% across Paytm's core consumer flows — via SSR migration, route-level code splitting, and edge caching. The approach was adopted as the performance standard across all product verticals.
The frontend platform I own — rendering contracts, module boundaries, shared infrastructure — serves Paytm's entire consumer product suite at India scale, every single month, without fanfare.
Seven years of the same platform growing in scope and complexity. From writing the first reusable components to owning the architecture that 8 product teams depend on today.