ModelSpec is a declarative, infrastructure-aware specification for managing AI models across their full operational lifecycle — from deployment and scaling to observability, cost, and compliance.

This guide is written for: ML engineers deploying models to production, MLOps / Platform engineers, supporting LLM workloads, Technical leads, reviewing or standardizing AI deployments

ModelSpec is designed to be useful before, during, and after model deployment. This page outlines common situations where ModelSpec adds immediate value — even if you only adopt a small part of it.

ModelSpec is a declarative specification for describing AI models in production — not just the model itself, but the intent, constraints, and assumptions around how it should run.

PIQC helps teams understand whether their AI models are correctly, efficiently, and safely deployed in production, by reasoning over what is running, how it is configured, and what it is expected to do.