better decisions.
with a named owner for each one.
kint is a German company building decision intelligence for operators. Planners get better plans for the decisions they make every day, and kint keeps models, data and runs in one place. Changes go through review before they run.
- start2grow 2025/26
- Top 4 of the competition.
- Backed by startUP campus HFU
- Startup programme of Hochschule Furtwangen University.
- Startup program
- kint is a member of NVIDIA Inception.
research, architecture and AI engineering.
Chief Executive OfficerProf. Dr. Pascal Laube
Leads strategy and product. Professor for Software Systems with a PhD in Machine Learning, with experience in optimization and in leading engineering teams.
LinkedIn
Chief Technology OfficerPedro Castelo Branco Lourenço
Leads technology and infrastructure. More than 20 years of enterprise software architecture. Microsoft MVP and Distinguished Engineer.
LinkedIn
Lead AI EngineerJozef Mikyta
Drives AI development and the model training pipeline: the systems that turn data and plain language into optimization models.
LinkedInCompanies sit on mountains of data. Every day their teams make hundreds of decisions based on experience, spreadsheets, and gut feeling. Models, data and workflows sit in separate departments, steps are moved by hand, and nobody can trace which result led to which decision. They know better answers exist in their data. They just can't get to them.
kint changes that. We build a model of how the operation runs and play through many scenarios. We show the best plan the solver finds within the stated rules, including options that are hard to spot by hand. Planners still plan, and dispatchers still dispatch. They do it with better options and the reasons in view.
AI builds the model. The math backs the decision.
better decisions
kint's decision agent turns how planners describe their work into an optimization model. kint's forward deployed engineers (FDEs), who build each solution with the customer, review it with them. Established methods find the best plan within the rules and report how close to the optimum it is.
accountable decisions
Every model, data source and run in one place: what exists, where documentation or oversight is missing, who released which decision. Changes go through review before they run. The decision agent proposes, and a named person on the customer's side releases each plan.
four things changed at once.
Optimization has worked for decades. It stayed with the few organisations that could afford a team to build and maintain the models. That is changing.
- 01 · language models
plain language becomes a formal model
Language models can now draft an optimization model from the way planners describe a decision. The expert's job moves from writing the model to checking it.
- 02 · sprawl
models everywhere, overview nowhere
Decisions run on models built by different teams and suppliers over many years. Few organisations can say which ones exist, what they use and who answers for them.
- 03 · regulation
documentation and human oversight
The EU AI Act asks high-risk AI systems for record keeping and human oversight. Decisions with high impact need evidence and an owner, not only a result.
- 04 · people
specialists are scarce
Operations research specialists are scarce, and many organisations have none in-house. Their time should go to review, where it counts most.
six rules we hold ourselves to.
They shape the product and how we work with customers.
- 01
amplify people
kint makes planners stronger. The planner still plans, the dispatcher still dispatches, with better options.
- 02
show the reasons
A plan from an exact model names the limit that binds it, so a team can explain the decision.
- 03
show it
When kint calls a plan optimal, the solver's bound and gap show it.
- 04
connect, don't copy
kint sits next to existing systems, and nothing gets ripped out. It keeps versioned copies of the exports a decision uses, and nothing more.
- 05
a named owner per decision
Each decision has a named owner, a version and a record. Nothing changes in production without review.
- 06
established science
Established, tested methods. New ideas earn their place by evidence.
what kint brings.
kint builds the software, runs it for you in Germany and works with your experts from the first model on.
- decision engineOptimization with the solver's bound and gap.
- the agentBuilds models with planners, in plain language.
- decision managementEvery decision model, run and change in one place, with gaps in documentation and oversight shown.
- the recordEach decision is versioned, can be re-checked against its recorded inputs and is released by a named person.
- FDEskint's FDEs review each model with domain experts, from the pilot on.
- flexibilityOne engine takes any decision with rules, goals and data: a new model, the same engine, the same record.
talk to the people who build it.
No sales layer: you speak with the founders and kint's FDEs.
- Research
- Konstanz · Furtwangen
- Operations
- Hosted in Germany, operated by kint
- Contact
- hello@kint.pro