Last month Snorkel AI highlighted how we’re deepening our partnership with Microsoft Azure AI to help enterprises and government agencies solve their most impactful problems. One example of how Snorkel AI helps organizations leverage Azure AI services for proprietary data and custom objectives is the new Snorkel Flow integration for Microsoft Azure Form Recognizer (currently in private preview).
Azure Form Recognizer is an AI service that provides pre-built and customizable models for analyzing forms and PDFs. In addition to pre-built models supporting standard forms like W-2s, invoices, receipts, business cards, etc., Form Recognizer includes custom training support to fine-tune its powerful neural document models and support proprietary datasets, non-standard formats, and custom objectives.
Check out this demo to see how Snorkel Flow can be used to easily and quickly train Azure Form Recognizer custom models. In this demo, we showcase the integration with a real-world use case from real estate and construction. Companies in this space manage massive volumes of forms and contracts with high variability depending on the year or specific process being followed. Without Snorkel Flow, it would require significant manual effort to extract specific key dollar amounts (such as the original contract amount, the current contract amount, and any contract value modifications included in the document) from a large data set of real estate and construction contracts.
The demo also shows how Snorkel Flow can be used to orchestrate document content preprocessing (including OCR, layout information, and more through Form Recognizer), kick off custom Form Recognizer training jobs directly from the UI, and auto-generate performance analyses over custom Form Recognizer models to guide your next steps.
We’re excited to continue deepening our partnership with Microsoft Azure AI to accelerate AI development for enterprises. Schedule a custom demo tailored to your use case and Azure stack with our ML experts today.
Recommended articles
View all articles
OSWorld 2.0: Why Long-Horizon Computer-Use Agents Still Fail Four Out of Five Tasks
Mengqi Yuan (XLANG Lab, University of Hong Kong) presents OSWorld 2.0, a benchmark of 108 long-horizon, real-world computer-use workflows where even frontier AI agents complete only 20.6% of tasks outright after 300+ steps each.
September 3, 2026
•
Snorkel Team
Fable 5.1 on Frontier Coding Tasks: Efficient Successes, Distinct Failure Modes
We evaluated Fable 5.1 on a series of frontier coding tasks from our proprietary Terminal-Bench+ dataset and compared the results against Opus 5. Fable remained competitive across most categories and was materially more efficient on successful runs, while its gap was concentrated in a small set of terminal-heavy and build/dependency tasks. Because category sizes are small and uneven, we treat
September 1, 2026
•
,
Terminal-Bench 4.0: Why Continuous Benchmarks Require Continuous QA
The speed of new frontier model releases keeps accelerating. Meanwhile benchmarks struggle to keep up and saturate quickly, often being left in the dust. Most benchmarks are static datasets with no active maintenance, causing them to lose value fast. Some benchmarks are looking to change this by becoming Continuous Benchmarks. Terminal-Bench is one of the most widely reported benchmarks on
August 28, 2026
•
