Building infrastructure for AI systems.
Currently curious about observability, distributed systems, and making debugging a little less painful.
I'm 20, based in Delhi, and I spend most of my time building backend systems and the infrastructure beneath them.
Most projects begin the same way: I run into a problem I can't stop thinking about, and building is usually how I figure out the answer.
Right now that question is:
What does observability look like when the system you're observing is itself making decisions?
🔍 VOIDAI agents are moving into production, but debugging them still feels like guesswork. VOID instruments agent executions, surfaces abnormal behaviour, and leaves enough context behind that engineers can understand why something failed—not just that it failed. Stack
|
Most AI code review starts with raw Git diffs. PullShark cleans and prioritizes changes before they ever reach a model, letting the LLM spend its context understanding code instead of metadata. Stack
|
I enjoy contributing to software that already has people depending on it.
Most of my work has been in the Talawa ecosystem under the Palisadoes Foundation, where I've spent time fixing concurrency bugs, improving backend performance, tightening authorization, and building tooling that makes the project easier to maintain.
Working in someone else's codebase has taught me that understanding why code exists is often harder—and more valuable—than writing new code.
→ Explore the projects:
I write occasionally—not tutorials, just notes from things I've learned while building.
📝 Building VOID: Finding Why AI Agents Fail
More articles → medium.com/@bhatiayug175
When I'm not building, I'm usually reading about distributed systems, browsing engineering postmortems, or trying to understand why production systems fail in interesting ways.