Neur.on. A Swiss company that builds its own models for Swiss law

We are a team based in Fribourg, Switzerland. We develop LexMachina, an engine trained on Swiss legal and financial data, and Corrext, the platform where legal experts put it to work. We publish no client names and no testimonials: anything we claim should be verifiable somewhere other than our own website.

Request a demo

What we can prove

A young company that works with lawyers and banks must accept being judged on facts that can be checked elsewhere. Here are ours: awards from the Swiss ecosystem, a research project supported by the Confederation, and a certification issued by an independent body.

You will find nothing else on this page: no client logos, no quotes, no testimonials. Anything not listed here is something we cannot demonstrate publicly.

2024

  • Swiss FinTech AwardsWinner, Early Stage categoryAward
  • BILANZMost innovative companies in Switzerland, ranked 13thAward
  • ISO 27001Information security certified in February 2024, Swiss hosting with InfomaniakCertification
  • UBS Female Founder AwardFinalist among ten candidates, a distinction for the founderAward

2023

  • Digital ShapersBILANZ, a distinction for the founderAward
  • Le Temps · Forum des 100The founder among the 100 personalities of French-speaking SwitzerlandAward

R&D

  • InnosuisseInnovation project supported by the Confederation, carried out with the iCoSys institute of the School of Engineering and Architecture of Fribourg (HEIA-FR)Innovation

“Technology Empowers, Not Overpowers”

Our guiding principle fits in one sentence, and it shapes everything else. A legal translation carries responsibility: that of the legal expert who signs it, not that of a model. So we build tools that augment legal professionals. They give them the context, the official terminology and the alternatives they need to decide quickly, and they leave the decision to them. An engine that decided in their place would not be progress: it would be a transfer of risk.

That is why we never predict the end of the translator or the lawyer-linguist. The four quality levels below put this principle into practice: you choose the share of machine and the share of human work that the stakes justify, and you always know which one you chose.

The AI proposes, the legal expert decides

Alternatives compared, terminology verifiable in context, harmonisation suggested: never an automatic approval.

The official source is shown

CHnell displays where a term comes from, in which domain and in which text, so that the choice can be justified.

Control stays with your organisation

Engine policy, access rights, Highly sensitive content mode: the rules are yours, and the tool applies them.

See the Corrext platform

Four quality levels, because not every document carries the same stakes

You choose the human share document by document. It is the most concrete form our principle takes.

  1. 1

    Machine translation

    For quick understanding: document review, internal reading, sorting a case file. The engine alone, without review, and you know it.

  2. 2

    Post-editing

    Light or Full review: contracts, General Terms and Conditions, correspondence. A lawyer-linguist reviews the text so that it can be published.

  3. 3

    Legal review

    For high stakes: legal opinions, publications, binding documents. The review is carried out by experts who take responsibility for it.

  4. 4

    Certification and apostille

    Certified and legalised translation, through a network of expert translators, for official files.

What is Swiss, exactly

“Swiss solution” is a phrase many providers use. Here is what it covers in our case, point by point, so that you can compare it with any other offer.

The entity
Neur.on AI Solutions SA, a company under Swiss law based at Place de la Gare 15, 1700 Fribourg. No parent company outside Switzerland holds our clients' data.
The infrastructure
Hosting with Infomaniak, in Switzerland. Your documents are stored on Swiss servers in every case, beyond the reach of the US CLOUD Act, in line with the revised FADP and the GDPR. ISO 27001 certification obtained in February 2024.
The engine
LexMachina is trained on Swiss legal and financial data: Federal Supreme Court decisions (BGE/ATF), texts of the Swiss Financial Market Supervisory Authority FINMA, codes. Over 15 million multilingual segments, not a general-purpose model with a glossary added on top.
The processing
Exclusively in Switzerland with LexMachina or in Highly sensitive content mode. Third-party engines, hosted abroad, remain available, but they trigger the Attorney-Client privilege notice, which reminds you of your organisation's policy.
Quadrilingualism
Native German, French, Italian and English in every pair, and 30 languages on the platform. Federal law is authoritative in German, French and Italian, and international matters are handled in English: a Swiss legal tool cannot afford to treat any of these languages as secondary.
The research
An innovation project supported by Innosuisse, carried out with the iCoSys institute of HEIA-FR. Applied research stays in Switzerland too.
The terminology
CHnell indexes official Swiss sources across 30 domains, from legislation to financial publications. It also takes in your own resources, without mixing them with those of another client.

The Corrext platformRequest a demo

An expert team in Swiss LegalTech

Neur.on brings together lawyer-linguists, natural language processing engineers and data scientists in Fribourg. Some know what a term commits you to in court. Others know how to teach it to a machine.

Applied research is carried out with the iCoSys institute of the School of Engineering and Architecture of Fribourg (HEIA-FR), as part of a project supported by Innosuisse, the Swiss Innovation Agency.

Paula Reichenberg

Founder

A lawyer who studied law in Fribourg, she holds a lawyer's practising certificate from the Canton of Zurich and an MBA from INSEAD. She also completed an executive programme in data science at EPFL. She created the first neural machine translation engine designed for Swiss lawyers, then founded Neur.on. BILANZ named her a Digital Shaper in 2023, Le Temps listed her the same year among the 100 personalities of French-speaking Switzerland, and she was a finalist for the UBS Female Founder Award in 2024.

Fabien Bernier

Chief Executive Officer (CEO)

He leads Neur.on and its teams, from strategy to the development of Corrext and LexMachina for legal and financial professionals.

Lawyer-linguists and engineers

Lawyer-linguists and natural language processing engineers work together, and with our clients, to keep improving the engine and the platform for the specific needs of Swiss law, tax and finance.

The kinds of organisations that trust us

We can say which professions work with us. We will not say which organisations. Our clients' confidentiality comes before our marketing: a law firm translating arbitration exhibits has no interest in seeing its name on its provider's website, and a bank even less so.

A named reference will never be published here without written consent. Until then, we prefer an honest list of categories to a flattering wall of logos.

Major Swiss firms

Financial institutions

Cantonal and federal authorities

Big 4

Legal publishers

Integration of a Swiss legal translation engine via API, or on-premises deployment.

No logos, no names, no testimonials on this site. If you would like to speak to an organisation comparable to yours, let us know: we will ask for its consent, and we will name it only if it agrees.

Discuss your case

Judge us on your own documents

Request a demo