Comet

Comet is an end-to-end model evaluation and observability platform designed for teams building and operating machine learning and LLM-based systems. It centralizes experiment tracking, model evaluation, and production monitoring so teams can understand, compare, and improve models across the full lifecycle. With Comet, users can log experiments from popular ML frameworks, capture hyperparameters, metrics, artifacts, and code, and organize them in a searchable workspace for reproducibility and collaboration.

The platform offers specialized LLM evaluation capabilities, including prompt and response logging, qualitative and quantitative evaluation workflows, and support for human and automated feedback. Teams can define custom evaluation metrics, run structured evaluations across models and prompts, and compare performance over time.