Case Study: Faraday Lab achieves faster, GDPR-compliant LLM training with Amazon Web Services

Faraday Lab Deploys LLMs Designed for French Accessibility in Days Using Amazon SageMaker

Faraday Lab needed a fast, cost-effective way to train and deploy large language models for French accessibility while staying compliant with GDPR. Using Amazon Web Services’ Amazon SageMaker, the company worked with open-source models from Hugging Face and aimed to improve model performance without requiring a large budget or long training cycles.

Amazon Web Services helped Faraday Lab cut LLM training time by 50%, bringing training down to just 2 days, and improve inference speed by 30%. The ARIA 70B version 3 model reached a multitask language understanding score of 64.75, and Faraday’s first six months saw 10,000 model downloads. By hosting endpoints in Europe, AWS also helped keep prompts and responses in the EU, supporting GDPR compliance.

Faraday

William Elong

Chief Executive Officer