Production Evals For Agentic AI Systems - Nishant Gupta, Meta Superintelligence Labs
- Source
- youtube.com
- Author
- AI Engineer
- Date
Context
An needs to establish whether the whole task succeeded, including its interactions with other software. In this production-evaluation talk, a software engineering tech lead at Meta proposes testing realistic workflows for tool failures, recovery, completion time and cost alongside answer correctness.
The proposed feedback loop turns observed failures into future tests. Teams collect execution records and user outcomes, have people review ambiguous cases, then add validated scenarios to the offline evaluation set. This gives a later model or tool update a concrete test of whether it reintroduces a problem users have already encountered.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
- determinism — Whether the same input reliably produces the same output — something LLM systems mostly lack, which changes how you test and debug them.
“The question is no longer did the model generate the right answer? The question is did the system behave correctly?”
Nishant Gupta
“Because benchmarks measure model capability. Production measures system behavior.”
Nishant Gupta
“Reliability becomes the North Star metric. Accuracy becomes the only input.”
Nishant Gupta
“Production traffic is no longer just traffic. It becomes evaluation data.”
Nishant Gupta
“Evaluation is no longer just a phase, it's an operational capability.”
Nishant Gupta
videoBetter AI Models Won't Fix Your Engineering Team — Nicholas Arcolano, Jellyfish
videoWhy AI Writing Still Reads Like Slop — Louis-François Bouchard, Towards AI
videoHow to Test AI Agents When No Benchmark Exists — Nathanael Weill, Air Canada
video1,368 AI Failures and What They Teach About Evals — Vishal Srivastava & Tanmay Sah, New York Life
Checking sign-in…
Loading comments…