Solutions / Sovereign and on-premise AI
For data that has to stay in Switzerland: open-weight models running where you decide.
These scenarios are illustrations. None of them describes a client.
Banking and finance
What we would build
A screening assistant that groups hits by the person they actually describe, using the birth dates, places and employers mentioned in each article. Analysts review one card per real person with the matching evidence highlighted, and every dismissal is logged with its reason.
Python, spaCy, Mistral, PostgreSQL
Banking and finance
What we would build
An assistant that answers advisors' questions from the bank's own research and product documents, quoting the exact paragraph each answer comes from. It only searches documents the advisor is already allowed to open, and it runs on the bank's servers.
FAISS, vLLM, Llama, Keycloak
Watchmaking and luxury
What we would build
An inspection station that photographs each dial under fixed lighting and highlights areas that differ from approved parts. Inspectors confirm or overrule each flag, and those decisions become the reference for the next model update.
PyTorch, Anomalib, ONNX Runtime
Watchmaking and luxury
What we would build
A briefing assistant that reads the client's purchase history, service records and advisor notes, then writes a one-page brief in the advisor's language with a source for every line. It runs on infrastructure the house controls and stores nothing of its own.
Mistral, vLLM, Salesforce API
Pharma and life sciences
What we would build
An intake pipeline that extracts reporter, patient, product, event and seriousness from each report, proposes MedDRA terms and highlights the source sentence for every field. Processors review the pre-filled case before it enters the safety database through E2B(R3) import, and the pipeline is validated for that intended use.
Mistral, MedDRA, Python
Pharma and life sciences
What we would build
A search assistant over your submissions, SOPs and Swissmedic, EMA and ICH guidelines that answers in a few sentences and links each claim to the exact section. It is read-only, and it declines to answer when the documents do not support one.
FAISS, bge-reranker-v2-m3, vLLM, Veeva Vault
Sovereign AI means the models, the data and the logs stay under your control. We deploy open-weight language and vision models on your servers or with a Swiss-owned hosting provider, sized for your workload, so no prompt or document is sent to a foreign API. You keep the model version fixed for as long as your validation requires.
For focused tasks such as extraction, classification and answering from documents, current open-weight models perform well, while the largest hosted models are still ahead for open-ended reasoning. We test open-weight candidates on your data during the feasibility study before recommending hardware.
It depends on model size and traffic. Many workloads run on one or two recent GPUs. We size it during the study and work with your IT team or a Swiss data centre.
Keeping processing in Switzerland and under your control simplifies compliance with the revised Federal Act on Data Protection, because no cross-border transfer takes place.
Your team or ours, as you prefer. We deliver monitoring, update procedures and documentation so the system can run without us.
We start with a feasibility study on your own data, so you see results before anything is promised.