Heidelberg startup kausable has closed a €12 million seed round to accelerate development of AI models that adapt to changing conditions without continuous retraining. The round was led by UVC Partners and Entourage, with participation from HTGF (High-Tech Gründerfonds), Mätch VC and angel investors connected to Black Forest Labs, OpenAI, Google DeepMind, Noxtua and the European Laboratory for Learning and Intelligent Systems (ELLIS).
A seed raise to build a causal “world model”
kausable is building what it calls a world model: a foundational representation of systems that emphasizes causal reasoning over purely statistical learning from historical data. The company says these models can infer relationships, adapt from limited information and solve problems not seen during initial training—reducing the need to continually collect massive datasets and retrain models.
TipPFN: zero‑shot forecasting for rare, high‑impact events
One of the flagship developments is TipPFN, a zero‑shot forecasting model designed to anticipate rare but high‑impact events in complex systems. Zero‑shot models attempt tasks without specific retraining for the scenario, aligning with kausable’s goal of making AI more robust to novel situations.
Academic validation and collaborations
Researchers at kausable recently co‑authored a paper with experts from Columbia University supporting the causal‑reasoning architecture behind their tech. While the details of the paper were not released, the academic collaboration bolsters the scientific credibility of their approach.
Founding team and operational plans
Founded in 2025 by Johannes Haux (CEO), Dr. Benjamin Herdeanu (CTO) and Gregor Ramien (COO), the three founders are physicists with ties to the University of Heidelberg and Black Forest Labs, and experience in regulated sectors such as banking and cybersecurity. kausable previously raised €1.5 million in pre‑seed funding at launch. The new funding will allow the company to expand its nine‑person team and continue developing what it describes as its “rapid‑learning frontier” model.
Why this matters for industry
kausable is targeting markets where conditions change quickly and rare events can have outsized consequences—robotics, energy, finance and healthcare, among others. By using models that infer causality and learn from small amounts of new data, the company aims to cut the cost and complexity of AI projects that today require frequent retraining cycles.
Investor and founder comments
“With kausable, we’re solving this problem by developing a new type of world model that learns efficiently and adapts robustly,” said Johannes Haux. Andreas Unseld, partner at UVC Partners, added that the solution turns AI “from a series of costly, one‑off projects into something that can be deployed at industrial scale.” Pieterjan Bouten, cofounder of Entourage, highlighted the methodological shift: rather than endlessly increasing datasets and retraining, kausable is pursuing a fundamentally different approach centered on adaptability and causal inference.
The new financing positions kausable to continue bridging research and industrial applications, with immediate plans to scale its technical capabilities and grow the team.
