Most factories demo well but fall apart the moment real work runs through them in production. This class covers what it takes to run a software factory a team can actually depend on: how to measure its effectiveness, what control flow keeps it recoverable, how to see inside it when it breaks, and why narrow scope plus your own data beats a general-purpose factory every time. You will leave with the building blocks to implement one against your own codebase.
Speaker:
Ash Tilawat, CTO, Gauntlet AI
This program is for software developers, ML/AI engineers and hands-on technical practitioners who actively work on production code. Please bring your machine to get the most out of this workshop.
This event is NOT meant for beginners or those who are new to software development.
What we cover
Evals for the factory. Scoring a pipeline, not a single model call. What to measure, what to ignore, and how to catch regressions before your users do.
Loops, harnesses, and orchestrator building blocks. The control flow that makes agents recoverable: retry, reconcile, verify. Where to put the human.
Narrow focus and training data. Constrained scope and company-specific data outperform generality, and how to pick the first slice worth automating.