AI beyond manual labeling
now supercharged by foundation models

Accelerate time to value with our transformative approach to data-centric AI—powered by programmatic labeling and now, foundation models.

Request a demo
Image
Image
Image
Image
Image
Image
Image
Image
Image
Image
Image
Image
Image
Image
Image
Image

Introducing Data-centric Foundation Model Development

New Snorkel Flow capabilities for enterprises to unlock complex, performance-critical use cases with GPT-3, RoBERTa, T5, and other foundation models.
Image

Pixability distilled knowledge from foundation models and built smaller classification models with more than 90% accuracy in days with Data-centric Foundation Model Development in Snorkel Flow.

Learn more

Snorkel Flow's data-centric, programmatic workflow speeds AI development by 10-100x.


Label faster

Current problem
AI requires large, high-quality training data sets, but labeling by hand is slow and expensive. Too many AI projects never get off the ground.
How it works
Label massive training data sets in minutes, not months. Write labeling functions to programmatically capture human insight and existing resources.

Snorkel Flow then applies and intelligently aggregates these labeling functions to auto-label millions of data points at computer speed.

Model faster

Current problem
Model errors that delay production often originate from the training data. Fixing the data is slow and difficult—and you’re flying blind as to what corrections to make.
Snorkel Flow solution
As you label training data, Snorkel Flow trains models in real-time, providing actionable guidance to get to production-grade performance.
With programmatic labeling, data iteration is remarkably efficient: simply edit or add labeling functions to address errors.

Adapt faster

Current problem
Once deployed, it’s difficult to adapt applications to real-world data drift or objective changes. Maintaining models can require complete manual relabeling.
Snorkel Flow solution
Package for production with a click, then adapt applications quickly with simple edits to label schema and labeling functions.
Snorkel Flow regenerates your entire training set so you’re ready to retrain your model in minutes (and stay in production).
Current problem
AI requires large, high-quality training data sets, but labeling by hand is slow and expensive. Too many AI projects never get off the ground.
Snorkel Flow solution

Label massive training data sets in minutes, not months. Write labeling functions to programmatically capture human insight and existing resources.

Snorkel Flow then applies and intelligently aggregates these labeling functions to auto-label millions of data points at computer speed.

Current problem
Model errors that delay production often originate from the training data. Fixing the data is slow and difficult—and you’re flying blind as to what corrections to make.
Snorkel Flow solution

With programmatic labeling, data iteration is efficient: simply edit or add labeling functions to fix errors and speed time-to-performance.

As you label training data, Snorkel Flow trains models in real-time, providing actionable guidance to get to production-grade performance.

Current problem
Once deployed, it’s difficult to adapt applications to real-world data drift or objective changes. Maintaining models can require complete manual relabeling.
Snorkel Flow solution

Package for production with a click, then adapt applications quickly with simple edits to label schema and labeling functions.

Snorkel Flow regenerates your entire training set so you’re ready to retrain your model in minutes (and stay in production).

Image

Gartner Research:
Cool Vendors in AI Core
Technologies 2022

Image
Image

CB Insights AI 100:
The most promising artificial
intelligence startups of 2022

Image
Image

Enterprise Tech 30:
by WingVC and Nasdaq

Image
Image

Data50: The World’s Top Data Startups

Image
Image

Madrona Venture Group and Goldman Sachs: Intelligent Applications Top 40

Image

The data-centric
AI development platform

Snorkel Flow gives you maximum data utility so you can create training data exponentially faster, iterate and adapt with ease, and ship more AI applications.

  • Use foundation models

    Supercharge AI development by adapting foundation models to your domain faster than ever, and by encoding domain-specific foundation model knowledge into smaller, high-accuracy deployable production models.

  • Label programmatically

    Easily create labeling functions rather than labeling data points one-by-one. Snorkel Flow uses these to auto-label vast training datasets in minutes.

  • Model instantly

    Snorkel Flow continuously trains and analyzes models to guide targeted iteration. You can also use the Python SDK to train custom models.

  • Iterate rapidly

    Go from analysis to action and reach performance goals quickly with pre-scriptive guidance to iterate on both models and data.

Deep platform support for complex ML tasks and data types

Build powerful AI solutions to handle real-world tasks over complex data types. Combine ML tasks for multi-model applications across a range of complex data types and formats.

Image

Structured data classification

Image

Sequence tagging

Image

Supported data types


Image
Conversational text
Image
Text documents
Image

Native PDFs

Image

HTML files

Image

Semi-structured/ tabular data

Image
Numeric data
Image
Network data
Image
And more


Get more from domain expert partnership.

Don't limit your subject matter experts' participation to tediously labeling one by one. Empower them to transfer their knowledge in a fraction of the time required by manual labeling. Activate their expertise for dramatically better training data creation, iteration, and troubleshooting.

Enterprise ready, fully interoperable

Cloud-agnostic and fully interoperable with your existing ML stack via an extensive Python SDK and other endpoints. Snorkel Flow provides enterprise-grade security and governance, user-tailored workflows, and access to unparalleled expertise.

Data ingest

Quickly and securely integrate to data pipelines or upload data locally.

ImageImageImageImageImage

Model training

Train custom models or choose from leading model frameworks with optional AutoML.

ImageImageImageImageImage

Production serving

Deploy your models within Snorkel Flow or export to the service of your choice.

ImageImageImageImage

Infrastructure

Host Snorkel Flow within the secure infrastructure of your choice.

ImageImageImageImageImage

Dive in

[get_press_posts]
Press
Blog
Research
Case studies
Press
Image
November 17, 2022
Snorkel AI Accelerates Foundation Model Adoption with Data-centric AI


Image
November 17, 2022
AI startup Snorkel preps a new kind of expert for enterprise AI


Image
November 17, 2022
Snorkel dives into data labeling and foundation AI models


Image
July 28, 2022
Here’s why a gold rush of NLP startups is about to arrive


Blog
Image
November 17, 2022
Data-centric Foundation Model Development: Bridging the gap between foundation models and enterprise AI


Image
November 17, 2022
Better not bigger: How to get GPT-3 quality at 0.1% the cost


Image
November 3, 2022
Building an NLP application to analyze ESG factors in Earnings Calls using Snorkel Flow


Image
August 4, 2022
The Future of Data-Centric AI 2022 day 1 highlights


Research
Image
2022
Universalizing Weak Supervision


Image
2021
Ontology-driven weak supervision for clinical entity classification in electronic health records


Image
2017
Rapid Training Data Creation with Weak Supervision


Image
2016
Data Programming: Creating Large Datasets Quickly


Customer Stories
Image
September 30, 2022
How Schlumberger uses Snorkel Flow to enhance proactive well management


Image
September 30, 2022
How a global custodial bank automated KYC verification with Snorkel Flow


Image
September 28, 2022
How Memorial Sloan Kettering Cancer Center used Snorkel Flow to scale clinical trial screening


Image
February 26, 2022
How Genentech extracted information for clinical trial analytics with Snorkel Flow


Image

Are you ready to dive in?

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