FileAI is an AI-native unstructured data processing platform specializing in automating document processing workflows with AI. Here is how we helped FileAI with the adoption of Generative AI.
Case study
FileAI processes almost half a billion documents per year, with products spanning markets across the US, Asia, EU, and India. One of their most important operations when automating document processing is data extraction and data entry. As documents present one of the most complex unstructured data types, FileAI is always looking for ways to improve the performance of their solutions. As Generative AI positioned itself as a leading technology for unstructured data processing, FileAI collaborated with us to further improve their AI document processing capabilities.
The project started with the goal of improving the two key data extraction and data entry operations: optical character recognition and form filling. Optical character recognition (OCR) is the process of converting documents into machine-readable text, which is used for all downstream operations. One of those operations is form filling, which refers to locating and extracting specific data from a document into a structured form for the purpose of data entry.
For both tasks the proposed solution combined traditional ML systems and Generative AI. We developed and trained LLM and VLM models coupled with text embedding models that drastically increased OCR and form-filling accuracy on complex documents spanning foreign languages and alphabets. The spotlight of the project was enabling FileAI to additionally obtain a semantic representation of the documents, such as was needed when documents contained visual elements like images or graphs.
For the supporting AWS infrastructure we used the following AWS services: EC2, ECS, S3, RDS, CodePipeline, Lambda, Bedrock, and SageMaker. Scalability was supported through AWS autoscaling. Additionally, a monitoring solution was developed that allowed FileAI to see all AI model calls live.
The solution now serves the production workload of FileAI and benefits its customers, and we continue to collaborate with FileAI on new AI technologies.
Working with Brando and his team has been fantastic for FileAI. He possesses deep knowledge of the AWS cloud platform and has a progressive stance on AI applications, specifically generative AI. We are continuing to engage his team for future projects to support FileAI’s mission.
Christian Schneider, CEO, FileAI