Professional secrecy
Law firms, clinics and gestorías handle data the law won’t let them move out of their jurisdiction’s control.
AI software · Built on NVIDIA GPUs
Sovereign AI for Spain.
Seats is a private AI assistant for Spanish law firms, gestorías and clinics, built on open models we fine-tune for Spain’s languages and run on NVIDIA GPUs under Spanish jurisdiction.
01 · The problem
Spanish firms can’t put client data into AI outside their jurisdiction.
“Data in Spain” isn’t enough
Jurisdiction follows the company, not the building: a US provider in Madrid is still subject to the CLOUD Act.
The result
They either don’t use AI, or staff use it quietly in personal tools. Both cost them clients.
62 %of Spanish companies want to run their infrastructure in a single country
90 %consider data sovereignty essential
Source: Nutanix, May 2026 (via Computing.es).
02 · The product
Seats: your firm’s AI, under Spanish jurisdiction.
One private assistant your whole team uses every day. The software your firm already runs can reach the same models through API access on the Galerna platform.
Seats
In developmentA private AI assistant for law firms, gestorías (tax and accounting advisers) and clinics: chat, document Q&A and drafting.
€30–60 per seat per month · planned list price
Included in the platform
API access
Gateway builtConnect the software your firm already uses to the same private models: OpenAI- and Ollama-compatible endpoints, on a fixed monthly plan.
- NVIDIA GPUs in Spain
- Spanish-owned company
- Zero retention of prompts and answers
- Signed residency declaration on every response
We sell software seats, not GPU time or tokens.Planned list prices. Seats opens first to pilot firms in Madrid.
03 · How it works
If we can’t prove where and how a model runs, we don’t serve it.
Your firm
- Seats
- API access
Galerna gateway
- Per-firm keys
- Zero retention
- Usage metering
Attested registry
- Real served precision
- Signed residency: ES
NVIDIA GPUs in Spain
- vLLM on CUDA
- Open models + Galerna fine-tunes
Built
OpenAI/Ollama-compatible gateway, prefill and decode metering, and verification that fails closed. 84 automated tests; end-to-end smoke test against a simulated vLLM backend.
Next
First deployment on a rented NVIDIA GPU in Spain, then billing, sign-up and TLS.
04 · NVIDIA in our stack
Every Galerna model runs on NVIDIA GPUs.
Revenue comes from software subscriptions. GPUs are our cost of goods, not our product: we don’t resell compute.
Inference
Open-weight LLMs (27B class) served with vLLM on CUDA. Target card: NVIDIA RTX PRO 6000, 96 GB, rented in Spanish data centres.
Fine-tuning
Fine-tunes of the best open models for Spanish, Catalan, Basque and Galician, trained on NVIDIA GPUs.
Attestation
Today: signed GPU-host declarations. Next: hardware-rooted proof with NVIDIA confidential computing on Blackwell-class GPUs.
CUDAvLLMNVIDIA RTX PRO 6000 · 96 GBEvaluating: TensorRT-LLM · NIM · NeMoRoadmap: confidential computing
05 · Galerna models
Spanish, Catalan, Basque and Galician, natively.
We fine-tune the best open models for Spain’s Spanish and its three other official languages, which is what public administrations and firms ask for.
EspañolSpanish, as written in Spain
CatalàCatalan
EuskaraBasque
GalegoGalician
- Public benchmark (planned) against ALIA, Salamandra and Qwen.
- ALIA has no official API: today nobody offers the co-official languages with residency guarantees.
- Open-weight models in the 27B class, served with vLLM on NVIDIA GPUs.
06 · Where we are
The core software is built. First customers are next.
Inference gateway
OpenAI/Ollama-compatible, per-firm keys, zero retention, usage metering, attested precision and residency. 84 tests pass; end-to-end smoke test on a simulated GPU backend.
Honest gaps
No production GPU yet. The Seats assistant is in development. Billing, sign-up and TLS are pending.
First customers
- First NVIDIA GPU live in Spain
- 3–5 paying firms on Seats
- Public benchmark published
We already have relationships with law firms that want Seats.
07 · Company
Two full-time co-founders. Both build the product and work with customers.
Company
Galerna is a commercial brand of Unisphere Overseas S.L.
Based in
Madrid, Spain
Product
Seats, a private AI assistant for firms, with API access
Customers
Law firms, gestorías and clinics in Spain; public sector later
Stage
Pre-seed. Inference gateway built; first GPU deployment next
Industry
Generative AI · legal and professional services · Spanish-language NLP
Contact
Your AI, under Spanish jurisdiction.
We’re starting with law firms, gestorías and clinics in Madrid. Tell us about your firm and we’ll show you Seats as it takes shape.
Alanalan@galerna.si
Jorgejorge@galerna.si