Snorkel AI | Solutions

Industry Solutions —

Use Snorkel Flow to build powerful AI applications

Finance

Document & contract intelligence
News and social media monitoring
Customer support & routing
Anti-money laundering (AML)

Government

Analyst triage
Named entity & event recognition
Relation extraction

Telecommunications & Cyber

Network traffic classification
Network anomaly detection
Cyber threat detection

Insurance

PDF form & multi-modal data processing
Fraud detection
Claims management & routing

Healthcare

Medical record extraction
Medical device monitoring
Physician assist & triage

E-commerce

Policy enforcement
Product cataloging
Fraud & spam detection

Case Study: Top U.S. Bank

A top U.S. bank uses Snorkel Flow to quickly build AI applications that classify and extract information from their documents. On one recent time-sensitive use case, the bank had estimated over a month of hand-labeling to build a model. With Snorkel Flow, the team produced a solution that was over 99% accurate in under 24 hours, and that could be quickly and easily adapted to new problems and business lines.

99.1%

Snorkel Flow Accuracy

< 24hrs

From problem start

> 250K

# Documents processed

Use cases —

A flexible platform for developing a range of use cases

Classification

Label and tag text, structured, semi-structured, and multi-modal data

Document classification
Fraud & spam detection
Sentiment analysis
Anomaly detection
Topic modeling

Information extraction

Extract entities, relationships, and structured information from complex documents and forms

Contract & document intelligence
Table, form, and PDF extraction
Named entity & event recognition
Knowledge management & discovery

Ranking

Ranking of content for relevance and other custom factors

Content recommendation
Analyst triage
Queue prioritization

Case Study: Proven Technology

Snorkel Flow is based on technology developed at the Stanford AI Lab, and deployed with some of the world's leading organizations. For example, the Snorkel AI team deployed an early version of the Snorkel Flow technology with teams at Google, finding that it could replace hand-labeling hundreds of thousands of data points and offer increased flexibility and speed.

10-100K+

Hand labels replaced

52%

Improvement by repurposing organizational resources

6M+

Labels in < 30 min.

Roles —

A platform for cross-functional collaboration

Successful AI solutions require collaboration across roles and teams — from the subject matter expert's knowledge to the ML engineer's deployment pipeline, Snorkel Flow is the only end-to-end platform that truly brings all roles together via the data.

Data Scientists

Accelerate your data science development process. Go fast and deep in developing complex machine learning applications quickly and with full access to modern, state-of-the-art techniques.

Build and manage training data and models with rich programmatic input
Use both GUIs and managed notebooks for speed and flexibility
Powerful data visualization, analysis, and monitoring tools

Subject Matter Experts

Go far beyond hand-labeling data. Inject rich domain expertise and drive end-to-end AI solutions.

End-to-end push-button GUIs
Expressive programmatic data labeling and management operations
Powerful data viewing, error analysis, and export capabilities

ML Engineers

Break free of the hand-labeled training data bottleneck. Rapidly and iteratively develop training datasets and then train, deploy, and monitor pre-installed or custom models.

Rapidly and iteratively develop, manage, and version training datasets and models
Push-button AutoML for model training & hyperparameter selection
Built-in model zoo or full Python SDK connectivity for custom model integration

Roadmap —

From research to production

Snorkel Flow builds on and productionizes years of research represented in over thirty peer-reviewed research papers and deployments with some of the world’s leading organizations, covering use cases in text, structured, semi-structured, telemetry, image, video, time series, and other data types.