Daxte Creates LegalMap, One of the First French Generative AI Solutions for Legal Documents Using AWS

Daxte was founded in 2022 and created a generative AI solution, LegalMap, that helps search for information in French legal documents. To develop AI models for its solution, the company fine-tuned open-source large language models such as Llama2 7b from Hugging Face with its own datasets and used Amazon SageMaker, a fully managed service that brings together a broad set of tools to provide high-performance, low-cost machine learning. “Using Amazon SageMaker, we can manage, develop, test, and deploy our AI models efficiently,” says Arnaud de La Bédoyère, founder of Daxte.

LegalMap then needed to scan 30 years’ worth of legal documents, digitizing them for the AI models to use. To accurately scan documents, the company used Amazon Textract, a machine learning service that automatically extracts text, handwriting, layout elements, and data from scanned documents. The company chose Amazon Textract because it provides a complete solution, a highly responsive team, and multifaceted support.

As of December 2023, LegalMap had a dataset of nearly 125 million documents comprising legal, financial, and marketing data. When it prepared the data for generative AI, LegalMap changed names, addresses, and other sensitive information to remain GDPR compliant.

In 2023, Daxte worked with Data Reply France, an AWS Partner, to optimize training costs by migrating from GPU resources to AWS Trainium, a high-performance machine learning training accelerator, and AWS Inferentia, which provides accelerators designed to deliver high performance at low cost for deep learning inference applications.

“Using AWS, we access cutting-edge technical expertise and robust resources,” says La Bédoyère. “This will result in smooth and efficient migrations of our complex infrastructure.”

Daxte is launching LegalMap in September 2024, with its first service—called Watch—for real-time information monitoring and alerts for legal firms across France. The company estimates that LegalMap will scan 600 million documents by 2026, and as its database grows Daxte will continue to use AWS to refine its generative AI capabilities. “The completeness of AWS services was a game changer for us,” says Arnaud de La Bédoyère. “Using AWS, we can serve our customers effectively and run our solution reliably.”