Case Study

Deploying LLMs designed for language accessibility with AWS and Data Reply

With the support of Data Reply, the startup Faraday Lab has created its own generative AI solution trained on French language datasets to improve accessibility and diversity.

Contact us

PRIVACY

I declare that I have read and fully understood the Privacy Notice and I hereby express my consent to the processing of my personal data by Reply SpA for marketing purposes, in particular to receive promotional and commercial communications or information regarding company events or webinars, using automated contact means (e.g. SMS, MMS, fax, email and web applications) or traditional methods (e.g. phone calls and paper mail).

Something went wrong with the form submit. Try again.

#AI training
#GDPR compliance  
#LLMs

Picture

The challenge

Reduce LLM training times without compromising data sovereignty or GDPR compliance in a cost-effective way.

the scenario

Balancing AI innovation with data compliance in the EU

Faraday Lab is a French startup specialising in the development and implementation of innovative solutions in the fields of artificial intelligence, machine learning, and data analytics. Situated at the intersection of innovative AI development and strict EU data privacy regulations, Faraday Lab aspired to refine their AI solutions. The prolonged training times of LLMs and the necessity for GDPR compliance presented a formidable challenge.

The company aimed to improve generative AI accessibility and diversity by training and fine-tuning LLMs on French language datasets, ensuring all could access without geographic limitations.

Picture

the solution

AI advancement with AWS SageMaker

The transformative breakthrough came as Faraday Lab partnered with Data Reply, leveraging their deep expertise in AWS services to optimise LLM training on Amazon SageMaker. Data Reply's guidance was pivotal in helping Faraday Lab halve their model training time and enhance inference speeds.

This synergy leveraged AWS SageMaker to ensure GDPR compliance, with data securely stored within the EU, demonstrating Faraday Lab's commitment to data sovereignty.

How we did it

Strategic expertise and
streamlined Machine Learning

With the support of Data Reply, Faraday Lab implemented AWS SageMaker to refine their training process, reducing it from several weeks to just 15 days using Hugging Face's open-source models. Data Reply's role was crucial in selecting high-quality, GDPR-compliant French language datasets, ensuring the AI's cultural acuity.

The collaboration extended to the deployment of Ares Chat and a bespoke Chrome extension, enhancing user interaction without sacrificing privacy or data integrity.

the results

Data Reply-enhanced AI training:
a model of GDPR-fidelity

Thanks to the collaboration with Data Reply and the utilisation of AWS SageMaker, Faraday has established a new paradigm in training efficiency and privacy adherence, resulting in a robust AI model that meets the intricate demands of GDPR.

Picture

50% faster LLM training time

Picture

30% increased inference speed

Picture

Improved accuracy of responses

Picture

GDPR compliance

Picture

Budget constraints respected

Data Reply

Picture

Data Reply is a Reply Group company and a leading AWS partner, offering a wide range of advanced AI, ML and data processing analytics services. The company operate across a range of industries and business functions, enabling customers to achieve meaningful business outcomes through the effective use of data, accelerating innovation and time to value.