--- title: "CSIDIA — AI agents on your company’s ontology, and private AI with open-weight models" description: "We build for your company a system that brings people, documents, case files, processes and data into a single ontology, governed and connected by AI agents that work for people with rules, permissions and confirmations. For sensitive data, open-weight models trained or fine-tuned on your task, on our NVIDIA infrastructure or on-premise at your site." url: https://csidia.com/en/ lang: en publisher: CSIDIA S.r.l. (https://csidia.com/) --- AI agents and open-weight models # One system for your company, governed by AI agents. With open-weight models trained on your work. We bring the company’s people, documents, case files, processes and data into a single ontology, that is a shared model of the company. AI agents govern and connect it, and work for people: within their rules and permissions, with their confirmation before the actions that matter. For sensitive data we use open-weight models trained for your purpose, on our infrastructure or yours. [Book a meeting](https://csidia.com/en/book/) [The system and the agents](https://csidia.com/en/what-we-do/#system) The system ## A single company ontology, governed and connected by AI agents. In a company, information sits in many programs, each with its own names. We build a shared model: every person, document, case file, process and piece of data has one name, one place and its links to the rest. AI agents work on top of it: they read, connect, prepare the work and carry out the permitted actions, on behalf of people. - people - documents - case files - processes - data For one person ### A personal assistant that coordinates a system of agents It coordinates one person’s work: one agent per project, mail, calendar, WhatsApp, the work queue and the log. The actions that matter start with the person’s click. For the clients of a firm ### The agent of each client on a professional firm’s platform Every person and every client organisation of the firm has its own agent. It works within that person’s boundaries and permissions and carries out the platform’s actions with their confirmation. For an organisation ### An agent that runs an organisation’s day-to-day case files Inside the organisation’s platform it keeps track of compliance duties, mail, deadlines and documents. It proposes changes and does not make them on its own. How they govern it ### With people’s rules, permissions and confirmations. Every agent has written rules, works within the permissions of the person it works for and asks for confirmation before the actions that matter. Every step is logged. Each agent uses the model best suited to the task, open-weight when the data cannot leave. - rules - permissions - confirmations - log [The system and the agents, in detail](https://csidia.com/en/what-we-do/#system) Our philosophy ## With pay-per-use models, your knowledge leaves the company with every request. Large closed models are used over the internet and paid per use, in tokens: every piece of text they read and write has a price. An open-weight model, trained for your purpose, learns from you and runs on our infrastructure or yours. That is why, when the data cannot leave, the agents too work with an open-weight model. Pay per use ### A closed model, paid per use - **Your data leaves the company.** Every request sends your documents to systems run by a vendor you do not control. - **It does not know your work.** It has general knowledge but does not know your company: the context has to be supplied with every request. - **The cost grows with use.** The more you use it, the more you spend. Prices, models and terms are set by the vendor. - **Nothing stays with the company.** The model belongs to the vendor: what you build depends on them. Trained for you ### An open-weight model for your purpose - **Your data stays under control.** On our infrastructure or yours. We always tell you which model reads it and where it runs. - **It learns your work.** Trained on your data and for your purpose: on that task it works better than a general-purpose model. - **No per-use fees.** No per-request charge to an AI vendor: the model runs on our infrastructure or yours. - **The knowledge stays yours.** What the model learns from your data stays yours, and grows with the company. The value is no longer the software or the single feature, but the company’s knowledge: put in order in a shared model and learned by a model trained for your purpose, so that it grows the company. On broad tasks, such as writing code or answering anything, large general-purpose models stay ahead. We work on one purpose at a time, and we measure it. The infrastructure ## Compute of our own, and the ability to build it at your site. Software can be rewritten and features can be copied. A model trained for your purpose needs dedicated compute infrastructure instead, to train it, specialise it and run it every day. This is ours. GB300 An NVIDIA GB300 server The core of our compute. The large model runs here. 2 × 96 GB Two NVIDIA RTX PRO 6000 Blackwell Max-Q In one workstation, 96 GB of memory per card. This is where we train and specialise models. 180 billion The large model, open-weight About 180 billion parameters in total. It runs on the server and on the workstation too. 40 Requests at once, on the server The usable capacity, measured by us. At your site ### Infrastructure like ours, at your site. We design, build, install and manage machines like these for you, with everything they need. - network - cooling - power - security - updates [Our infrastructure, in detail](https://csidia.com/en/infrastructure/) Two routes ## On our infrastructure or yours. First route ### On our infrastructure We train the model for your purpose and run it on our servers and workstations. To start straight away, with no hardware investment. - It starts with a trial on your documents. - The model runs on our server and our workstation. - We always tell you which model reads your data. Second route ### Infrastructure of your own, designed and managed by us We design it around your workloads, build it, install it at your site and manage it. The model runs there, and the data stays in the company. - Network, cooling, power and security: we take care of them. - Updates and new models, verified on your cases before they are adopted. - The machines stay at your site. You can also start on our infrastructure and move to yours later. What we build on it ## Four examples of models at work. On this infrastructure we build concrete applications. Four capabilities already put to the test, with results we measured ourselves. They are examples: your purpose may be a different one. Documents and standards ### Checking technical documents against standards Every requirement searched for across the whole document, page by page, with the outcome and the citation. Your expert signs. In one of our trials **4,346** requirements across four machinery manuals, none skipped. How we do it → [https://csidia.com/en/capabilities/#standards](https://csidia.com/en/capabilities/#standards) Archives and documents ### Reading hard archives and documents Scans, drawings and tables turned into searchable data, without losing the codes. In one of our trials **98.4%** of codes and values on 1,174 pages of manuals from different manufacturers. The best tool used alone: 81.4%. How we do it → [https://csidia.com/en/capabilities/#archives](https://csidia.com/en/capabilities/#archives) Images and materials ### Generating and measuring custom images A model trained on your images generates new surfaces at true scale, pixel by pixel, and measures them against the real ones. Our measurements **520 dpi** on a full surface of about 3 by 2 metres, generated in an hour and a half. How we do it → [https://csidia.com/en/capabilities/#images](https://csidia.com/en/capabilities/#images) Data and mail ### An agent that prepares the work, a person who decides It reads mail and documents, keeps track of deadlines and case files and prepares replies and texts, with the model best suited to the task, open-weight when the data cannot leave. The rule No text goes out without a person’s approval. How we do it → [https://csidia.com/en/capabilities/#agent](https://csidia.com/en/capabilities/#agent) How it starts ## Three steps. You can stop after any of them. 1. **A meeting** You tell us the purpose. We tell you whether an agent or a trained model would do it better than you do today, what it would take (data, compute, time) and which route fits. It lasts 30 minutes and is free. 2. **A trial on your documents** On a sample of your documents or data, under a non-disclosure agreement, on our infrastructure. We measure the result with numbers you can check, on your cases. We agree it at the first meeting, tailored to you. 3. **At work** The model trained for your purpose runs on our infrastructure or yours; the agents work in your programs, with your permissions. They prepare the work, a person decides. We always tell you which model reads your data and where it runs. [How we work, step by step](https://csidia.com/en/how-we-work/) The first step ## Book a meeting. Tell us the purpose. We tell you whether an agent or a trained model would do it better than you do today, what it would take (data, compute, time) and which route fits: our infrastructure or yours. If it is not a job for us, we say so. - video call or in person - 30 minutes - free [Book a meeting](https://csidia.com/en/book/?servizio=sistema) [or write to engagement@csidia.com](mailto:engagement@csidia.com?subject=First%20meeting%3A%20A%20system%20of%20AI%20agents%20on%20your%20company%E2%80%99s%20ontology) The service A system of AI agents on your company’s ontology People, documents, case files, processes and data in a single ontology, governed and connected by AI agents that work for people, with rules, permissions and confirmations. [How it works](https://csidia.com/en/what-we-do/#system) The notes ## Real questions about AI in companies, with sources. New notes come out in Italian. The English archive holds the notes published up to October 2026. [The notes, in Italian](https://csidia.com/note-operative/) [English archive](https://csidia.com/en/notes/) Page: https://csidia.com/en/ · CSIDIA S.r.l., Vignola (MO), Italy · engagement@csidia.com · index for agents: https://csidia.com/llms.txt