The Evolution AI platform combines natural language processing and computer vision to extract data with up to 99.5% accuracy
The Evolution AI platform completely automates mundane back-office tasks. We remove the need to manually extract data from mountains of paperwork.
We use computer vision and natural language processing (NLP) to extract data from PDFs, scans, images and any machine-readable text. Our platform is pre-trained on 25 million documents so it learns new documents without needing substantial additional training.
The development of our award-winning platform was funded by one of the largest ever UK Government grants for AI R&D.
Our algorithms can learn to extract data from a new document type with only one example
Our platform can handle any document type, including PDFs and images.
English, Romanian, Japanese, it doesn’t matter. Our platform learns in any language.
Our algorithms are so powerful they can learn from a single example. In just two minutes you can create a new document type and start extracting user-defined fields.
Our algorithms prioritise learning from the most complex documents. This rapidly improves extraction from complex and simple documents simultaneously.
Active learning algorithms require far fewer training examples to learn a task. As illustrated, the active learning algorithm (blue) reaches the same accuracy with a third of the training examples compared to the same algorithm without active learning (orange).
Our models employ transfer learning, a technique where a model is pre-trained on a large dataset before attempting a new task. In Evolution AI’s case, the pre-training is on one of the largest stores of labelled documents ever assembled.
As shown here, the pre-trained model (blue) needs four times fewer training examples to reach the same accuracy, as the untrained model (orange). As the number of training examples increases, the pre-trained model remains consistently more accurate.
Our platform goes beyond processing PDFs and scanned documents. We use our award-winning NLP to extract, classify and match text in emails, websites or any other source.
Our information extraction capability was four times more successful than competitors at identifying businesses named in UK court listings. Suitable for: extracting company details, data from stock exchanges, dates of a contract, and more.
Our platform saves Dun & Bradstreet 100,000 hours of manual data processing a year by classifying millions of businesses into custom industry codes. Suitable for: classifying emails, screening negative news and tweets, and more.
Our matching capability avoids the overwhelming number of false-positives generated by text mining. Suitable for: finding related clauses in legal contracts, pairing related news articles, matching CVs to jobs, and more.
Evolution AI works with major UK banks, such as RBS and Natwest.
We offer interpretable models that can explain every decision made and provide a full audit trail. This allows any decision to be easily reviewed and challenged. This builds trust in our AI and provides assurance from a regulatory and governance standpoint.
The partnership with Evolution AI is significantly easing our joint customers’ requirements.
VP Product and Alliance Management, UiPath
RPA allows anyone to design a “robot” to perform repetitive tasks taken on a computer. These robots can then repeat the actions, trigger responses and communicate with other systems, automating repetitive tasks.
Any high-volume, rules-driven, repeatable process will benefit from RPA. We enable RPA to:
By delegating processes to a digital workforce of AI-enabled robots, employees can focus on more strategic and cognitive initiatives.
We offer different deployments to suit your company’s needs. Our platform works with an industry-standard HTTP REST API. Clients are able to post data at the endpoint and get the results in a documented JSON format.
Our most popular deployment is a service that is fully managed by us, with an infrastructure that is shared.
Private SaaS is a service that is fully managed by us on our hardware. The hardware and infrastructure are solely dedicated to your project.
We provide an Amazon Machine Image or a docker image for deployment on your own private cloud.
For our enterprise clients, we offer an on-premise deployment option and batch API to upload data using SFTP/FTPS and download the results.