Community GPU Grant Application for RAG-based AI on EU Privacy Law (EDPB & GDPR)

Dear Hugging Face Team,

We are two students and developers currently working on an open-source, Retrieval-Augmented Generation (RAG) based AI assistant specifically designed to support privacy professionals in navigating complex European data protection regulations.

Our project is trained on the complete archive of documents published by the European Data Protection Board (EDPB) — the EU body responsible for the consistent application of the General Data Protection Regulation (GDPR) across member states — along with the full legal text of the GDPR itself. This means our dataset comprises every single document ever published by the EDPB, including guidelines, decisions, and meeting minutes, all of which are critical resources for compliance and legal interpretation in the field of data protection.

For reference, the full EDPB document archive is publicly available here:
https://www.edpb.europa.eu/our-work-tools/documents/our-documents_en

Our goal is to provide a free and accessible AI assistant that enables privacy professionals, legal experts, and DPOs (Data Protection Officers) to quickly find relevant guidance and regulatory context without needing to manually search through thousands of PDF documents. We've already received highly encouraging feedback from members of the privacy and legal community who see strong potential in this tool.

Currently, we're in the development phase and are evaluating different models and vector search configurations to identify the best-performing architecture for our needs. One of our main efforts right now is building a robust local pipeline that preprocesses and structures PDF files for AI-friendly ingestion. This step includes automatic text extraction, intelligent chunking, metadata enrichment, and indexing — all optimized for downstream use in RAG pipelines.

Due to hardware limitations, we’re only able to work with a small subset (10–20) of the more than 1,800 documents we plan to use. We are therefore applying for a Hugging Face Community GPU Grant to help us scale our work. With access to a community GPU, we would be able to:

• Index the full document collection at scale
• Evaluate and fine-tune more powerful embedding and QA models
• Speed up model testing and improve overall accuracy and responsiveness
• Develop a user-ready experience

If granted access, we will adapt our implementation to make full and efficient use of the provided GPU resources.

Thank you for considering our application. If you have any further questions, we are glad to answer!

Warm regards,
Berkay, Arne