Skip to content
Neural Daily

BusinessNews

Open models in Europe's public sector, new persona data

At GTC Berlin, NVIDIA showcased European government projects running open models on infrastructure they control. For developers, the two new synthetic datasets stand out: one million U.K. and German persona records each.

By: Mikola Máté Patrik · · 2 min read

NVIDIA open models for European public sector projects
Image: NVIDIA · Press kit · Official NVIDIA press/blog image · Source

What happened?

On October 22, NVIDIA outlined where open models are running in European public administration (NVIDIA blog). The common thread: governments can evaluate models locally, adapt them to local languages, run them on their own infrastructure and keep data inside the organization.

New datasets for developers

  • Nemotron-Personas-UK: London-based Locai Labs released 1 million synthetic personas reflecting the U.K.'s demographic and geographic diversity. It helps test whether a model works consistently across age groups, education levels and urban or rural populations.
  • Nemotron-Personas-Germany: the Soofi research initiative's open dataset of 1 million records, each with six synthetic personas. It is grounded in the 2022 census across all 16 federal states and 400 districts, built with NeMo Data Designer and no real personal data.

Both work with NeMo User Sim, an open source library that evaluates and trains models and agents against simulated users in local languages. Polish and Slovenian editions are in the works with SpeakLeash and the University of Ljubljana.

Projects by country

  • European Commission: deepset is expanding the Haystack Enterprise Platform behind AI@EC, the Commission's AI platform on its own infrastructure, adding cost comparison, answer quality evaluation and tracing.
  • Poland: with Deloitte, the eHealth Centre is building a medical imaging platform on DGX BasePOD and MONAI. The labor ministry's agentic system is designed to support more than 400 labor offices using PLLuM, GPT OSS and Inkling models.
  • Germany: Materna's on-premises Sinfonie platform answers spoken questions from institutional knowledge. For classified data, secunet, NVIDIA, Red Hat and deepset are developing a sovereign cloud on the SINA Cloud foundation in German data centers.
  • France: public purchasing body CANUT selected a Twake.ai-based workspace from LINAGORA and Docaposte, available to more than 2,500 organizations without a separate tender.
  • Greece: KIEFER is building an agentic system on DGX SuperPOD with Greek-language tools.

Why it matters

For public sector AI, the question is rarely model quality and mostly where it runs and who sees the data. The synthetic persona datasets add something more concrete: bias and consistency testing without real citizen data. There is no Hungarian edition on the list yet.

    Produced with the help of AI, reviewed by an editor. How we make our articles

    Reviewed and edited by the Neural Daily desk. About

    Related articles