ML researchers and engineers who want a fast, self-hosted alternative to track training runs and compare metrics.

Aim is an open source experiment tracker that logs metrics, hyperparameters, and artifacts from training runs and offers a fast UI for comparing thousands of runs.

Self-hosted, running locally or on a team server, storing run data on disk without a required cloud account.

Used during model training and fine-tuning to track and visually compare large numbers of experiment runs.

It is built for speed at high run counts, a common pain point with heavier hosted tracking tools.

Integrates with PyTorch, Hugging Face Transformers, Keras, and other common training frameworks via lightweight SDK hooks.

Open source, Apache 2.0 licensed, self-hosted, actively maintained with over 6000 GitHub stars and commits as recent as August 2026.