AI infrastructure engineer

Sachin Malhotra

Anthropic · 2026

Sachin Malhotra worked on Anthropic’s Developer Infrastructure team at World’s Fair 2026, building and operating CI/CD systems. His work spans test reliability, CI observability and controls for agents with write access to production systems. He advocates operational budgets and infrastructure-enforced limits around agent autonomy. Before Anthropic, he held engineering roles at Etsy and Microsoft and earned an MS in computer science from the University of Southern California.

1 conference talk

AI Engineer World's Fair 202619:53

Give the Agent a Budget, Not a Token — Sachin Malhotra, Anthropic

Sachin Malhotra argues for bounding agent autonomy with operational budgets. Drawing on an accidental workload deletion incident and CI examples, he describes asymmetric permissions, replenishing rate limits, aggregate monitoring, and an undo test. Prompts explain intent, while infrastructure enforces limits and supplies trusted agent identity.

Sachin Malhotra

References

ai.engineer/talks/rbjWzZK2LU0-give-agent-budget-not-token ↗