Weights & Biases

Weights & Biases (W&B) is a widely adopted AI developer platform that serves as a system of record for training, fine-tuning, evaluating, and monitoring machine-learning models and LLM applications. Founded in 2017, it built its reputation on experiment tracking, letting data-science and ML teams log runs, metrics, hyperparameters, datasets, and artifacts for reproducibility and collaboration; more than a million practitioners and 1,400-plus organizations use it. Its governance-relevant capabilities center on W&B Registry, a curated central repository providing model and dataset versioning, aliases, lineage tracking, and access controls as a single source of truth for what is in production, and W&B Weave, an LLM observability layer that traces inputs, outputs, code, and metadata, ingests OpenTelemetry traces, integrates with major model providers, and supports online evaluations of production agents. The platform's governance posture is developer- and lifecycle-oriented rather than compliance-first: it provides the audit, lineage, and monitoring foundations GRC teams need but is designed primarily for ML engineers. It offers multi-tenant SaaS, isolated Dedicated Cloud, and customer-managed self-hosted deployments, and is certified under ISO 27001, SOC 2, and HIPAA.