Skip to content

Engineering, research and advisory

AI that works
in your industry.

brillQ is an applied AI company in Poland. We build AI systems for companies across industries. AI is our tool, not magic: people build the systems and answer for the result.

Services

What we do

Four areas of work, from the first idea to a system running in production.

  • AI strategy and consulting

    We help companies find where AI can solve a business problem and what it takes. A team of specialists and domain experts works with you for the length of the project, then disbands.

  • Executive AI advisory

    We advise executives and boards on where AI fits, what to build first and what to leave out, with a clear view of costs, risks and applicable rules such as the EU AI Act.

  • AI architecture and engineering

    Our engineers design the architecture of new AI systems and build them as software, together with the data they need. We integrate them into your existing environment and take them to production.

  • Applied AI R&D

    Researchers and engineers carry out applied research and experimental development in AI and advanced computing and in its use across industries. We build our own software libraries from this research and publish the results. Read about our applied AI research.

Why brillQ

AI as a tool, not magic

Many companies promise that AI can do everything. We use it where it helps, and we say when it does not.

Your problem first

We start from your data, your processes and your rules. If AI is not the answer, we tell you.

Tested before we rely on it

We test models and methods in our own applied AI research before they go into client work.

People build it and answer for it

Engineers, data scientists and researchers design, build and test every system. A person from our core team leads the project and answers for the result.

Built for production

We build systems to run in production, connected to what you already use. You set the goals and approve the key decisions.

Competences

AI competences we work in

Everything at brillQ starts with these competences. Every system we build, our R&D and our work for clients grow out of them.

Agentic AI

Software that takes a goal, plans the steps and uses your tools to carry them out. We build it with clear limits, and people approve what matters.

Routine multistep work done faster, inside limits you set.

Generative AI

Models that create new text, images, code and other content from a plain request. They give people a strong first draft to review and refine.

Faster first drafts in your own format and style.

Large language models

Language models read and write natural language. We adapt them to your terminology, so they can summarize, classify and extract information the way your teams do.

Less time spent reading, sorting and rewriting text.

Machine learning

The foundation under most AI: algorithms that learn from examples to classify, rank, detect or estimate. We choose the simplest method that does the job.

A model fitted to your data and your question.

FAQ

Frequently asked questions

What does brillQ do?

brillQ is an applied AI company in Poland. We advise companies on AI strategy, and we design, build and integrate AI systems for them. In our R&D Lab we carry out applied research and experimental development in AI and build our own software from it. People do the work, and one person from our core team answers for each project.

Our four areas of work

Which industries does brillQ work in?

brillQ works across industries. We start from your problem and your data, not from a model. Our R&D Lab works in seven research areas: energy, construction and buildings, health and medicine, space and aerospace, defense, robotics and autonomous systems, and AI and advanced computing. Four of them have published papers so far, with results in fields from battery storage and weather risk to public hospitals and health data.

See our results by field

Where is brillQ based, and does it work in Polish?

brillQ is a company registered in Poland, and our working languages are English and Polish. We take on projects for companies in Poland, across Europe and in English-speaking markets. Several of our papers are tested on Polish public data, such as day-ahead power prices, weather records and national grid data. To start a conversation, write to us in either language or open the Contact tab on any page.

What has brillQ published or built?

So far the brillQ R&D Lab has published 11 open papers, which anyone can read, and seven of them were tested on real data. From this research brillQ has built two systems. One software library is available now and one more is announced. More systems, libraries and applications will follow.

See brillQ in Numbers

What does the Q in brillQ stand for?

The Q in brillQ stands for the quaternary sector: the knowledge-based part of the economy, where the main output is new knowledge, methods and technology rather than raw materials, goods or routine services. Research and development is part of it, and so is the work of the brillQ R&D Lab. Every industry belongs to its own sector, but research and development for any of them is quaternary sector work.

What our R&D Lab does

Who does the work at brillQ?

People. At brillQ, AI systems are designed, built and tested by specialists: machine learning, data and software engineers, researchers and advisors, brought together into a team for each project and led by a person from our core team. The decisions and the accountability for the result stay with people.

How does brillQ use AI in its own work?

At brillQ, AI is a tool in our specialists' hands, not magic. They use AI where it helps, for example to search sources, analyze data or test code, and a person checks everything it contributes. When AI is not the right answer to a problem, we say so. The design, the engineering decisions and the accountability for every system stay with people.

What does an engagement with brillQ look like?

An engagement with brillQ runs in five steps. We define the problem, the goal and the data with you, and agree on the scope. We assemble the project team. The team builds and tests in short cycles and shows working results early. The project lead reviews every result and you make the key decisions. At the end you get a system in production or a clear recommendation.

What is a Liquid Enterprise?

Liquid Enterprise is the name brillQ uses for its operating model: an organization without permanent departments that takes the shape of each problem. A permanent core team leads every project. For each project a team of specialists is assembled, works for as long as the project lasts, then disbands. What brillQ's R&D learns is kept in its own software, systems and tools.

What is the core team at brillQ?

The core team is the permanent group of people at brillQ. They decide which problems we take on, scope projects with clients, assemble and lead the project teams, and answer for the results. Every project has a project lead from the core team.

What is a project team at brillQ?

A project team is a group of specialists that brillQ assembles for one problem: machine learning, data and software engineers, researchers or advisors, chosen for the skills that problem needs. It works with you for the length of the project and disbands once the work is handed over.

What is a project lead at brillQ?

At brillQ, the project lead is the person from the core team who leads your project. They stay your contact from the first meeting to the handover, review every result before it is delivered and are accountable for it. The key decisions stay with you.

What are the three layers of the AI systems brillQ designs?

brillQ designs AI systems in three layers, whatever their technical architecture. The first sets the goal and the limits with the client: what problem the system solves, what it may do on its own and which decisions stay with people. The second is orchestration, which splits the work across models and tools. The third, where a system uses them, is AI agents with defined roles and an audit log. Agents propose or, within the limits set with the client, act on their own, and people approve the decisions reserved to them at the agreed points.

What is decision architecture?

Decision architecture is the part of an AI system's design that settles who or what decides what. brillQ agrees on it with the client before the build: which decisions the system may take on its own, which it only recommends, which stay with people, and what data it may use. It is part of the system's design, not a project roadmap.

What is AI orchestration?

In the AI systems brillQ designs, orchestration is how a problem that needs more than one model is handled. The system divides it into tasks, assigns each to the model or tool suited to it, runs them in parallel or in sequence and merges the results into one consistent answer. It is a way of designing, fitted to each problem, not a separate product.

What is an agent swarm?

An agent swarm is a pattern inside some AI systems brillQ designs: a group of AI agents that the system forms for one task. Each agent has a defined role, and the group works in parallel within limits set with the client. When the task is done and a person has reviewed the output, the group is dissolved. It is a design pattern in the software, not a team at brillQ.

What is human review in an AI system?

In the AI systems brillQ designs, agents analyze and propose or act within set limits, and people verify and approve the decisions reserved to them. A person reviews the output at the points agreed with the client before the build. Each agent leaves a record of what it did in an audit log, so the way a result was produced can be traced. People answer for the result, not the agents.

How much can AI lower energy costs with battery storage?

In a brillQ R&D Lab simulation, an optimization-based AI agent planning a battery saved a median 9.8% of energy cost in the first year and 10.9% in the second, before battery wear, against rules that barely use the battery. A stochastic predictive controller saved the same, so the gain comes from optimization. With battery wear priced at 0.08 EUR/kWh, only a small net saving remained. The test replayed public Polish day-ahead prices with the recorded load of 20 buildings in Ireland and the UK and a modeled battery; when the agent behaved as intended, an independent guardian kept every site within its state-of-charge limits.

Read the paper on battery storage

How does brillQ design AI systems with the EU AI Act in mind?

brillQ designs AI systems with human approval steps, audit logs and written documentation, to help a client meet its own duties on oversight, record keeping and transparency. In the hybrid digital layer we built, language model agents have no write access to setpoints and every model deployment needs human approval. brillQ advises on the technical and organizational side of the Act but does not give legal advice or certify compliance. The client's compliance function decides.

See the hybrid digital layer

Can bias detection under Article 4a of the EU AI Act use synthetic or anonymized data?

Only partly. In brillQ R&D Lab tests on a simulation and on three real credit datasets, how well an alternative matched an audit with the true protected attribute depended more on the fairness measure and the group size than on the method. Synthetic data from published aggregates fails for outcome-based measures, because it has no real outcomes. Standard k-anonymity removes the rare minority cells that carry the strongest bias signals. Central differential privacy at epsilon = 1 comes closest to a true audit, but still processes sensitive data at a trusted third party. Not legal advice.

Read the paper on bias detection

How does brillQ handle personal data and GDPR requirements?

brillQ designs systems to use only the personal data a task needs and to keep it where the client controls it. In a research test on synthetic construction logs, we ran the open Polish language model Bielik locally on a CPU, without sending data to external services, and our research examines when synthetic or anonymized data is enough. Roles, lawful basis and transfers are for the client to assess with its data protection officer or legal adviser. brillQ does not give legal advice.

Read the paper on synthetic and anonymized data

How does brillQ approach DORA and NIS2?

DORA binds financial entities directly. NIS2, through national law, binds essential and important entities in covered sectors, in general those of medium size and above. Both require them to manage ICT provider risk, supply chain security and secure development. brillQ designs AI systems that fit into the client's existing environment, can run on local models and keep audit logs, so the client can cover them in its own management of ICT provider risk. Our energy reference architecture has a threat model mapped to the EU AI Act and NIS2. brillQ does not claim compliance or certification. The client's security and risk team decides what is sufficient.

Read the paper on the energy reference architecture

How do brillQ's AI systems fit an ISO/IEC 27001 security management system?

ISO/IEC 27001 certification covers an organization's information security management system, not a product. brillQ's own certification is in progress; it is not certified yet. brillQ designs AI systems that can run on local models, with audit logs and human approval, so they can sit inside a client's own management system and its supplier requirements. The client's security team decides which supplier evidence it needs.

See a system we built

How do I contact brillQ?

Write to hello@brillq.today or open the Contact tab on the right edge of any page, which opens a short contact form. Tell us what you are working on and where you would like to use AI, and brillQ will reply by email.