What Jua does
Jua builds foundation models for physical reality, learning physics directly from data to power AI systems that operate in the physical world. Its approach pairs a world model that predicts how physical systems behave with agents that act on those predictions, starting with atmospheric and energy applications.
Key capabilities
Jua's EPT-2 world model is trained on atmospheric data and produces state-of-the-art atmospheric prediction, and the same base model can be fine-tuned for other physical domains such as airfoils and shock waves. Its Athena agent takes an objective and a world model, simulates consequences, calls tools, and resolves the task. The technology supports atmospheric forecasting, prediction-market and derivatives pricing, and research acceleration.
Who it's for
Jua serves energy traders and utilities, hedge funds, and large enterprises that need accurate physical-world forecasting and optimization. The company is based in Zurich, Switzerland, has published research at ICLR and NeurIPS, and lists clients including TotalEnergies, Shell, Enel, and EDF.