“Mateusz showed a high level of expertise in designing AI systems, from hardware architecture through model selection to implementation. He could explain complex technical issues precisely and clearly to people without specialist IT knowledge.”

Tell us what you want to change. We check what is possible and build a system that works every day.
What should change and how we will know it worked.
Security check: We agree which data is sensitive and who may see it.
We gather data, tools and limits in our AI audit app.
Security check: We start with the tools you already have. We do not add another one unless it is needed.
A quick version we test together with you.
Security check: We test on safe or anonymised data.
We check that it works and pays off. You see progress in HQ.
Security check: We connect to your systems and your sign-in.
Data flows on its own and your team gets a working tool.
Security check: Data stays where you decide. We monitor the system continuously.
New opportunities
We measure the result against the metric from step 1. A working system opens new opportunities: from them we pick the next goal and take the same path again.
Documents
Safe search across company documents. Data does not leave the company.
AI agent team
Our team of AI agents that writes, reviews and tests code.
Finance
Reads the news about listed companies every day and turns it into a digest.
Healthcare
Automatic reading of qEEG recordings. Research stage.
Projects and tasks
The command centre for our projects. Tasks, sprints and an assistant that answers with sources.
A quick start in the Warsaw region, sign-in with Google accounts, scaling without servers of your own.
Google holds ISO/IEC 27001, 27017, 27018 and 27701 certificates and a SOC 2 report. By default the system runs on our account and Google processes the data as our subprocessor. If you prefer, the account is yours.
For companies that work in Microsoft 365. The Poland Central region in Warsaw, the AI model in the EU data zone.
Microsoft holds ISO/IEC 27001, 27017, 27018 and 27701 certificates and a SOC 2 report. By default the system runs on our account and Microsoft processes the data as our subprocessor. If you prefer, the account is yours.
A local AI model and fully closed infrastructure, offline if you need it. Data never leaves your network, and you remain its controller.
You are the data controller. The AI model runs on your server, so documents and questions never reach outside services. Without your consent we have no access.
Open a provision to see what it says, how we carry it out and where to read it.
What the provision says: The controller implements technical and organisational measures, such as pseudonymisation, to apply principles like data minimisation, and by default processes only the data necessary for each purpose.
Ours: before a document reaches the model, the system can mask national ID numbers, tax IDs, bank accounts, e-mail addresses and phone numbers. We also commit that your documents are used only for search and that we never train models on them.
What the provision says: The controller and the processor implement measures appropriate to the risk. Examples include encryption and the ability to ensure the confidentiality, integrity and availability of systems.
Ours: encrypted connections and document content, the database and sign-in on a separate network, access only to documents you have rights to, and a session that ends after 30 minutes of inactivity.
What the provision says: The controller is responsible for complying with the data-processing principles and must be able to demonstrate that compliance.
The event log helps with that: it records sign-ins, document views, AI questions, permission changes and exports (who, what, when, from where), and who looked at the log. For an audit you export it to a file.
What the provision says: The provision requires the confidentiality of systems and that only people acting on the controller's instructions have access to personal data.
Ours: sign-in and permissions are managed by your own systems, for example a Google or Microsoft account with two-step verification. If you have none, we deliver one with two-factor sign-in: you get full access and you run it. Two-factor sign-in is our practice, not a requirement of the provision.
What the provision says: A provider of an AI system that interacts directly with people ensures they know they are dealing with AI, unless that is obvious. Paragraph 2 also covers marking synthetic outputs in a machine-readable format.
We support this: an answer in the interface is marked as AI-generated. It also names its source (your document, general knowledge or out of scope), and questions that call for a decision are passed back to a person.
What the provision says: Processing on a controller's behalf requires a contract (art. 28). Where high risk is likely, the controller carries out an impact assessment before processing (art. 35).
Ours: before deployment we check with you whether an impact assessment (DPIA, carried out by the controller) is needed and how the system is classified under the AI Act. When we have access to the data, in the cloud on our account or during support, we sign a data processing agreement.
GDPR art. 28 (EUR-Lex) (opens in a new tab)GDPR art. 35 (EUR-Lex) (opens in a new tab)
The summaries of the provisions are ours and do not replace the text of the legal act. This is not legal advice.
From people we have built artificial intelligence systems with.
20 minutes. Tell us what you want to change and we will tell you honestly whether data can help.
Book a 20 minute call (opens the booking window)