Snorkel

Snorkel provides expert data development as a service for AI labs and enterprises building specialized models and agents. Its Snorkel Data Series offers ready-to-use datasets and agent environments, alongside custom benchmarks and training data. Deliverables include execution harnesses, verifiers and reward modules, letting teams train and evaluate agents against measurable tasks. Snorkel Solutions develops custom enterprise agents, with engagements covering deployment, monitoring and iteration. The business centers on expert services rather than a self-serve labeling application.

Founded out of the Stanford AI Lab in 2019, Snorkel is led by co-founder and CEO Alex Ratner, who started and led the earlier open-source project. The project's 2017 research described a way to create training data through weak supervision: users write labeling functions that encode heuristics, and Snorkel denoises their potentially inaccurate, correlated outputs without ground-truth labels. This lets domain knowledge contribute to training data without requiring people to label every example manually.

Snorkel serves specialized workflows in software engineering, science, finance, healthcare and other domains. In 2025, the company reported working with seven of the top ten US banks, alongside Fortune 500 companies and federal agencies. Customers can purchase non-exclusive Data Series products or commission custom development, including work with proprietary data. Snorkel supports existing model endpoints and compares models against workload performance and cost.