Build from the failure case
I start by asking how a system loses context, calls the wrong tool, or becomes impossible to debug. The happy path comes after the boundaries are clear.
Profile / 00
I design AI systems that can explain themselves.
My work sits between agent architecture, product engineering, and open-source reliability. I turn uncertain research ideas into systems people can inspect, operate, and improve.
Current focus
Autonomous memory, multi-agent coordination, retrieval quality, evaluation gates, and safe tool execution.
I’m especially interested in the moment a prototype stops being a clever loop and starts becoming dependable infrastructure.
AIPrimary domain
OSSPublic workshop
IndiaWorking globally
Keep autonomy observable
Long-running agents need explicit state, replayable decisions, and human-readable traces. Autonomy without an audit trail is just hidden complexity.
Ship proof, not promises
A project earns confidence through tests, measurable behavior, and a clear explanation of trade-offs—not through a polished demo alone.