Snorkel Flow

The first AI platform that supports the complete machine learning lifecycle, no hand-labeling needed.

How Snorkel Flow works —

Label data programmatically, train models efficiently, improve performance iteratively, and deploy applications rapidly—all in one platform. 

Rather than hand-labeling thousands of data points by hand, use Data Studio to programmatically label massive amounts of training data using labeling functions—rules, heuristics, and other custom complex operators—via a push-button UI or Python SDK using integrated notebooks. Get started quickly with ready-made labeling functions (LF) builders, data exploration tools,  or nifty auto-suggest features. Receive instant feedback with coverage and accuracy estimates of your LFs to develop a high-quality training data set.

Ready-made Labeling Function Builders

Snorkel Flow automatically learns the different labeling functions’ accuracies, denoises and integrates them, and stores versioned LF packages and training data in Data Manager. Unlike with hand-labeled data, you create training data in Snorkel Flow using code, so you can audit, modify, or serve it almost instantly. Snorkel Flow makes it easy to share resources, both LF and training data, with others on the team--no need to reinvent the wheel.

Train state-of-the-art ML models with a button push or via Python SDK using integrated notebooks to plug into your existing modeling pipelines. Snorkel Flow provides access to popular open-source model libraries that you can train on CPU- or GPU-based computing infrastructure. Snorkel Flow makes it easy to tune your models with automated hyperparameter search. Deploy these high-accuracy models immediately as real-time or batch APIs or via the SDK.

Snorkel Flow Model Zoo

Snorkel Flow includes several commonly used and custom analysis tools to compare multiple models over different data splits. It offers suggestions on improving model quality by adding or editing LFs or optimizing the model to target specific errors. You can also monitor performance drifts in LFs or the model and rapidly adapt to changes without relabeling from scratch. The result: AI development is now an iterative process rather than a one-and-done exercise that leaves performance on the table.

Features —

Advanced features to foster collaboration across roles, from data scientists and developers to subject matter experts, and leverage data at enterprise scale to build highly-accurate models.

Application Studio is a visual builder with pre-built solutions for industry-specific use cases and common AI tasks, giving you a head start developing ML-based applications over your data. Packaged application-specific pre- and post-processors, labeling functions, models, as well as a library of operator DAGs make customizing applications as easy as dragging and dropping new operators into the application flow. 

*Application Studio is in preview and will be generally available later in 2021.

Integrating Snorkel Flow with other machine learning and data systems is as easy as writing a line of Python–quickly integrate your existing training labels, data, models, and full applications with Snorkel Flow’s SDK and API access points at all stages in the development/deployment pipeline.

Snorkel’s technology has been proven to work with a wide range of data types and the use of cross-modal data--enabling solutions for use cases that weren’t possible before.

Snorkel Flow’s no-code GUI, Python SDK, and developer APIs are fully-interoperable, so your entire team—from engineers and data scientists to subject matter experts—can collaborate in the development workflow.

Range of enterprise-grade secure deployment options available including single-node or distributed on Kubernetes based on your needs.

User community —