Weights & Biases (often abbreviated W&B or wandb) was founded in 2017 by Lukas Biewald, Chris Van Pelt, and Shawn Lewis. Based in San Francisco, the company has raised over $250 million in funding, including a $50 million Series C in 2023, and is used by researchers and ML teams at thousands of organizations.
The platform started as an experiment tracking tool — a way to log hyperparameters, metrics, and model outputs during training runs and compare them visually. If you’ve ever trained multiple versions of a model and lost track of which configuration worked best, W&B solves that problem. Their tracking dashboard has become almost standard in the ML community.
Beyond experiment tracking, W&B now offers dataset versioning (Artifacts), model evaluation (Tables), hyperparameter optimization (Sweeps), collaborative reports, and model registry. The platform integrates with virtually every major ML framework — PyTorch, TensorFlow, Hugging Face, JAX, and more.
W&B is used by over 1,000 organizations, including OpenAI, Nvidia, Microsoft, Toyota, and many top academic research labs. When you read an AI research paper and see those neatly formatted training curves, there’s a good chance they were generated with W&B. The company has found the sweet spot between being developer-friendly enough for individual researchers and enterprise-ready enough for large ML teams managing hundreds of experiments and models simultaneously.
