Return to page

H2O Document AI

Extract Hidden Insights and Automate Document Processing

H2O Document AI automatically makes highly accurate AI models that classify documents, extract text, tables, and images, and group, label, and refine the extracted information. The service supports a wide variety of documents and use cases, helping organizations understand, process, and manage their large amounts of unstructured data.

Most organizations possess large numbers of documents, and some, such as patient health forms, are essential to day-to-day business processes. The others contain a vast reservoir of untapped insights, but in the past, it was nearly impossible to process and extract insights from all of these documents. H2O Document AI helps organizations more quickly and accurately process both business critical documents for increased productivity and other documents to find hidden insights.

How it Works

Document AI uses a combination of Intelligent Character Recognition (ICR), which leverages learning algorithms for generalizable character and word recognition, document layout understanding, and Natural Language Processing (NLP) to rapidly make highly accurate models.


Upload your documents to H2O Document AI using the Document AI web interface or API. Document AI allows you to handle a wide-variety of documents, including:

  • Image scans (faxes in PDF or other formats, pictures with text, and non-editable forms)

  • Documents with embedded text which have text and layout metadata (PDF docs, Word docs, HTML pages)

  • Documents with regular text “left to right/top to bottom” (CSVs, emails, editable forms)

Pre-process documents before training with a set of state-of-the-art computer vision and NLP product features.  Pre-processing includes support for

  • Recognizing and handling embedded text

  • Recognizing and handling logos

  • Page orientation resolution

  • Deskewing

  • Cropping

  • Text formatting optimization

  • Color binarization

  • Addressing input PDF quality challenges

Add, improve, and validate document labels.

  • Automatically creates labels for unlabeled documents

  • Automatically fixes data labeling errors in training data

  • Provides labeling interfaces for data scientists and 3rd party human labelers

  • Integrates with common label formats

  • Provides advanced options for validating labels against scored documents and determining labeling sufficiency

Select the training data set within H2O Document AI, and it will automatically learn the document and create models.

  • Language understanding and layout recognition using learning based on deep learning, transformer architectures, and machine learning

  • AI-ML engine that uses multiple computer vision and NLP algorithms for diverse AI tasks

    • Entity recognition

    • Document and page classification

    • Form understanding

    • Grouping & set identification

Post-process to ensure consistency, accuracy and organization of scored documents. H2O Document AI enables customers to perform a range of customized post-processing jobs that use AI algorithms vs. rules to ensure high quality predictions and insights.

  • Organizing prediction sets

  • Confidence and probability measures

  • Datatype standardization – date, times, currency codes, international numerical formats, locations

Publish models either to H2O MLOPS, which is part of the H2O AI Hybrid Cloud, or into your cloud or on-premises environment of choice.

Integrate Models into existing systems, processes, and applications via APIs or JSON documents.

Benefits of H2O Document AI

H2O Document AI is designed to help organizations automate document processes and find insights throughout their large volume of documents.

  • Frees up teams to do higher value work activities

  • Provides relief to users/analysts/managers by increasing efficiencies and reducing process redundancies

  • Allows the enterprise to go beyond OCR-based template methods and RPA-based memorization efforts which are not scalable as the variety and volume of documents changes.
  • Organizations can focus on quicker time to value for primary users and secondary consumers by focusing on developing applications and quicker integrations using highly accurate extracted information and knowledge (as opposed to updating rules, re-automating template management, and moving documents)


Automatically Learns and Trains Document AI Models

Select a data set with PDFs or images, bring your own labels or use our labeling workflows, and H2O Document AI will do the rest. We combine Intelligent Character Recognition (ICR) with Natural Language Processing (NLP) and layout intelligence, to rapidly make document management easier - pipelines, customized models, self-supervised learning, and process intelligence.

Generates Highly Accurate Results Fast

Make Document AI models without extensive time and rework to get the right results. Instead of just using optical character recognition (OCR), we designed H2O Document AI with our Kaggle Grandmasters and a diverse set of customers, to develop our combination of intelligent character recognition (ICR) with natural language processing (NLP)

Integrates with Existing Applications and Workflows

Ingesting documents, processing, training, and scoring are accessible via REST APIs, making the service easy to integrate with existing systems. Scoring outputs from H2O Document AI are flat JSON files that can also be accessed via an API.

Integrates with H2O MLOPS in the H2O AI Hybrid Cloud

Monitor and Manage Document AI Models with H2O MLOPS. MLOPS provides a monitored environment for scoring and also addresses A/B testing, logging, error handling, and scaling.

Integrates with H2O Wave for Custom Document AI Applications

Make AI apps with H2O Wave, an open-source and low-code Python development framework that makes it fast and easy for data scientists, machine learning engineers, and software developers to develop real-time interactive AI apps with sophisticated visualizations. With H2O Wave, data scientists and developers can rapidly make AI applications that include workflows and visualizations that make it easy to interpret and consume scored documents.

Featured Success Story

University of San Francisco (UCSF) Health is one of the top 10 hospitals in the United States and ranked #1 in neurology & neurosurgery.

UCSF Health was struggling with numerous document processes, such as medical referrals. Poor effectiveness from optical character recognition and robotic process automation solutions, led to wasted time and effort as well as poor customer experiences. With H2O Document AI, UCSF has been able to fully automate multiple processes for multiple different documents, leading to higher levels of efficiencies, higher value work, and better patient outcomes.

quotation mark

When we started this journey, we were hopeful that information extraction from semi-structured documents was possible, but we weren’t sure. Some in the industry told us it couldn’t be done. Now that the collaboration team has delivered, it opens up many possibilities.

Bob Rogers, Expert in Residence for AI, UCSF

Get Started

H2O Document AI is now available. Please fill out the form, and we’ll reach out to provide you with a demo and additional information.