Nursing homes & institutional catering
An AI assistant to generate balanced monthly menus, comply with Swiss nutritional standards, and preserve your head chef's expertise. Hosted 100% in Switzerland.
Planning a month of menus is a huge mental load — and knowledge that cannot be passed on
Your chef's expertise cannot be passed on.
The recipes live in a spreadsheet or a paper binder. The method, the substitutions, the touch — it is all in the chef's head. The day they leave, on holiday or for good, your kitchen loses years of expertise. And the quality of your menus depends solely on their experience.
Reconciling 4 audiences, special diets and the budget every month — alone.
Nursing-home residents (geriatric recommendations), staff, day care (paediatric age groups), home-delivered meals — each audience has its own nutritional constraints. Add to this individual diets (textures, allergens, diabetes), the budget, seasonality. A task no one can take on for you.
An AI assistant trained on your recipes, your audiences, your method
The tool proposes, the chef decides. Each suggestion is sourced and can be validated. The AI does not replace your chef — it frees up their time to focus on what really matters.
Trained on your recipe book
The assistant learns the recipes, the method and the know-how specific to your kitchen. It makes suggestions in your style, respecting your expertise — instead of inventing a generic one.
100% in Switzerland, data protected
Apertus model (EPFL + ETH Zurich + CSCS research) hosted in Crissier (VD). Residents' personal data anonymised. nLPD compliant. No data leaves the country.
Hosted by us or by you
Either accessed through a secure web interface (like ChatGPT, but dedicated to your establishment), or delivered as a server on your premises — standalone, with no Internet connection required.
Decision support with citations
The tool proposes, the chef decides. Each suggestion cites its sources (original recipe, nutritional reference). Systematic human validation, auditable traceability.
The interface, at a glance
Accessible from a browser, like ChatGPT — but dedicated to your kitchen, trained on your internal recipe book, hosted in Switzerland.
Illustrative preview — conversations and suggestions are fictitious, for demonstration purposes.
A nursing home in Vaud — pilot under way
An anonymised example from our active pilot in a medico-social establishment in French-speaking Switzerland, validating the approach under real conditions.
The context
Medico-social establishment in Vaud, 90 residents, central kitchen producing for four audiences: nursing-home residents, staff, an associated day care centre (several age groups) and home-delivered meals. Several years of internal recipes in Excel and expertise concentrated in the chef.
The deployment
AI assistant fine-tuned on the internal recipe book and the chef's method. Hosted on our infrastructure in Crissier — secure web interface, access for the chef and management. Pilot in 4 phases (scoping, preparation, specialisation, testing) over 5-6 months. No residents' personal data used for training.
The expected outcome
Reduced monthly planning time. Relevance of proposals across the 4 audiences. Optimisation of raw-material costs and waste. And above all: expertise now codified, transferable, no longer dependent on a single person.
Compatible with your obligations
The guarantees that make the platform usable in nursing homes and institutional catering with no compromise on confidentiality or nutritional safety.
Swiss nutritional standards — geriatric recommendations (nursing-home residents), paediatric (day care by age group), general public (staff and home-delivered meals). The assistant respects the ranges for each audience.
New Swiss data protection act (1 Sept. 2023). Residents' personal data anonymised. 100% hosting in Switzerland in Crissier (VD). No dependency on foreign cloud providers.
Tier IV datacenter in Crissier (VD), operated by 2itea. No foreign provider in the chain. On-premise mode available: the server stays in your building.
Stemming from public Swiss academic research (EPFL + ETH Zurich + Swiss National Supercomputing Centre). Apertus 8B base model (Sept. 2025), open and auditable code and weights.
Each proposal cites its source — original recipe from the internal book, nutritional reference. The chef validates. The AI never replaces human expertise.
You are the sole owner of your data and of the specialised model. Return or destruction on instruction, with certificate. No lock-in.
Six steps to move forward together
From the first demo to ongoing operations — a clear path to build a long-term partnership.