VEGA logo.

The AI engine that powers Affinda.

VEGA is an Intelligent Document Processing Engine

VEGA understands and processes any document type, making it possible to automate previously human-led workflows.
VEGA logo.
Can process any document type
"Human-in-the-loop" functionality
A range of powerful capabilities
World leading accuracy
Leverages the latest advancements in AI
Processes in 50+ languages

Vision, Language, Learning

The 3 pillars of the VEGA platform.

VEGA fuses these three capabilities into a single engine. Fusion of AI models is the key to intelligent document processing. Without this fusion, document processing projects fail.


Documents are designed for humans, by humans

Each document contains design elements for understanding. Understanding visual context is critical for accurate document processing. AI Models fail because they can’t ‘see’ the document. To understand all the information being presented, the AI needs to understand how a document’s layout conveys meaning.

What a human sees
a human readable invoice
What a machine sees
What a machine sees

Language capabilities across 50+ languages

Combining Vision and Learning with advances in Natural Language Processing:

VEGA can now automatically detect and process documents in 56 languages. It can also process multilingual documents – those documents that contain more than one language.

Using Transfer Learning, VEGA gets better with every document

VEGA is trained on millions of documents of many types, and can apply the lessons it learns to new challenges.

What this means:

Less data required: A new model can be trained with as little as 200 documents.

More affordable: Solving new challenges requires less time or resources.

Adaptability: VEGA can process new document types and designs that it has never seen before. This means it can work with smaller data sets than previously thought possible using AI.

Affinda is trusted by some of the world's leading organisations

VEGA is the reason our AI tech outperforms market leaders

8+ years of document IP
Latest Advancements in AI
Outperformance vs. competitors
Resume parser error rate
Resume parser error rate

Competitor 1 has been the market ‘leader’ for more than 15 years. Their core business is resume parsing

Invoice parser error rate
Invoice parser error rate.

Competitor 1 has a 30+ year history and over 1,000 employees. Their core business is information extraction.

AI now beats humans in document processing

A tidal shift is happening in document AI.

In 2019, researchers from Facebook, Google, and two universities developed a new benchmark for testing AI language understanding – SuperGLUE.

New AI models are tested against this benchmark to measure AI advances in language tasks.

A major breakthrough came in early 2021 with test results showing machines are now better than humans at language understanding tasks.

AI beats humans graph

Examples of VEGA in action.

Want to try the free tool?

Open this web page on a desktop device to upload your documents and see how it works for yourself.