AI-native architecture · Advisory
Most AI projects are theatre. I build the ones that aren't.
I help founders design AI systems and agents that actually ship — grounded in ROI and built to survive production, not to impress in a demo.
What I actually do
The gap is never the model. It's the architecture around it.
Architecture that survives production
Not a demo that impresses in a meeting. Systems built to run on real traffic, with real failure modes, long after the kickoff excitement wears off.
Agents you can trust to run unattended
Bounded autonomy, deterministic guardrails, and an evidence trail — so the system is verifiable at 2am, not just in the sandbox.
A team that keeps building after I leave
The goal isn't dependency on an advisor. It's a team that has absorbed the judgment, not just the deliverable.
Currently VP of Product & Engineering at Crossover. Former founding CTO at Recur Club, India's AI-native debt marketplace. Prior senior engineering roles at MakeMyTrip through hypergrowth, and at Airtel at national scale.
“Automation must be grounded in verifiability and ROI— not hype-driven trends.”
— from a published interview on building AI-native infrastructure
Is your AI plan real, or is it theatre?
A short conversation about what you're building, where the risk actually sits, and whether the architecture will hold once real traffic hits it.
30 minutes · no pitch · you'll leave with something useful either way