Theorema brings an experienced Czech founder back into startup life with an ambitious goal: help researchers choose better biological targets for new medicines. Its early results offer a reason to follow the company, while the next step is to establish how well those predictions work in practice.
Filip Doušek had already built and sold a technology company. He could have continued supporting other founders as an investor. Instead, the pace of progress in artificial intelligence drew him back into building a business himself.
In an interview with Czech technology publication Lupa, Doušek described that change of mind: investing no longer gave him the degree of involvement he wanted. The result is Theorema, founded with Hynek Walner. Lupa — interview with Filip Doušek
For readers following Czech startups, the company offers an interesting combination of entrepreneurial experience, international connections and a difficult scientific problem.
From business analytics to biology
Doušek previously founded Stories, an AI analytics company acquired by Workday in 2018. Workday described the technology as a way to automate analysis and give business users understandable insights into trends, problems and opportunities. Workday — Stories acquisition announcement
Theorema reunites experience from that chapter. Doušek is its founder and CEO; Walner, its co-founder and CTO, previously led machine-learning and AI-agent engineering at Workday. The company's scientific team includes specialists with backgrounds in target discovery, machine learning and experimental research. Theorema — founders and team
The new challenge is to help decide where a medicine should intervene in the biology of a disease.
Finding a useful point of intervention
A drug target is a biological component, such as a protein, that a treatment aims to influence. Choosing a useful target requires understanding whether changing it could affect the disease.
Theorema describes its technology as causal disease models. The aim is to predict the consequences of an intervention. Its own explanation makes an important distinction: something that changes alongside a disease may be responding to damage rather than causing it. An association alone does not establish a useful treatment target. Theorema — causal disease models
According to the company, its models combine evidence from sources including individual cells, genetics, experiments that disrupt genes and clinical studies. Its proposed applications include identifying targets, investigating other diseases in which an existing drug candidate might have a role, and examining which patient groups could benefit. Theorema — technology and applications
The commercial logic is straightforward: a research team wants stronger evidence before committing further work to a particular biological hypothesis. Theorema is building a product around that decision.
What the early results actually show
The company reports that an oncology benchmark recovered established relationships between cancer mutations and therapeutic targets 13.7 times more often than chance. This concerns a specific prediction task; it does not mean medicines can be developed 13.7 times faster. Theorema — oncology benchmark
In a separate test, its models predicted which pairs of gene disruptions a cancer cell could not survive. The company reports an AUROC of approximately 0.86 when both genes were withheld from training, compared with approximately 0.79 for the strongest published comparator shown on its website. AUROC measures how well a model ranks the two classes across decision thresholds; 0.86 should not be read as 86 percent accuracy. Theorema — published benchmark comparison
These are company-reported results. They provide a starting point for examining the technology, but do not establish that a proposed treatment is safe or effective in patients. Anyone considering a scientific collaboration should review the methods, data and experimental validation relevant to their own project.
AI also changes how the startup works
The research product is only part of the story. In the Lupa interview, Doušek also described a team working with several AI agents in parallel each day. He attributed much faster development to that way of working. Those statements describe the founder's experience, rather than an independently measured productivity result. Lupa — Theorema's development workflow
For other founders, this raises a useful operating question: where can AI shorten the path from an idea to something a customer can evaluate? In a research business, that also requires clear responsibility for reviewing evidence and testing outputs. Faster iteration becomes valuable when it helps the team reach a better-supported decision.
Czech research, international ambitions
Theorema lists a Czech entity in Prague and a US entity in Delaware. Its website names backers including Credo Ventures, KAYA VC, i&i Biotech Fund and Tilia Impact Ventures. Theorema — locations and investors
That structure gives international readers a concrete Czech startup to explore: a business connecting local research activity with an international commercial outlook. It also illustrates why evaluating a startup means looking separately at its scientific team, corporate structure, product and evidence.
The main conclusion
Theorema is worth following for the problem it has chosen and the experience behind the team. Its ambition is substantial, and the early benchmarks make the next stages of validation particularly interesting.
For potential partners, the practical next step is to define a specific research question and agree on the evidence needed to judge a pilot. For anyone watching Czech technology, this is a company whose progress can be assessed through the quality of its predictions and the experiments that follow them.
How Kodo can help
Kodo helps international technology and life-science companies prepare the commercial and operational side of entering Czechia. We support early market validation, partner mapping, local positioning, English and Czech communication, web content and coordination of the first implementation steps.
For legal, tax, regulatory, certification and investment-incentive matters, companies should work with qualified specialists and the responsible authorities.
Related reading
- Czech startups: sectors, funding and opportunities
- What Phonexia's South Korean acquisition reveals about Czech deep tech
- Finding business partners in the Czech Republic
Sources
- Lupa — Iva Brejlová's interview with Filip Doušek
- Workday — Stories acquisition announcement
- Theorema — founders, team, investors and locations
- Theorema — technology, applications and benchmark comparison
- Theorema — causal disease models and oncology benchmark
This article provides general business information as of 9 October 2026 and is not investment advice. Technical results are attributed to the company and should be assessed against the relevant methods and validation evidence.
