Jua

Jua is a frontier AI lab developing the Earth Physics Transformer (EPT), a weather model built on an end-to-end transformer architecture. Rather than solving physical equations sequentially across 3D grid cells and depending on precise initial conditions - the approach taken by traditional numerical weather prediction - EPT learns historical weather patterns and fundamental atmospheric physics directly. Its training combines observed weather data with physics-based simulations, including ERA5, to capture complex atmospheric dynamics at fast processing speeds.

Around the model, Jua offers a weather intelligence platform and APIs aimed at energy markets, where forecasts are used to support energy trading decisions. The stated emphasis is on consistent accuracy across the forecast horizons used in different trading products, and the company's applied focus spans energy trading, renewable energy and weather forecasting.

Technically, the work sits at the intersection of AI weather modeling, transformer architectures, atmospheric physics, physics-based simulation and physical AI. Hiring materials across roles emphasise this combination of data-driven AI with fundamental physics, along with a practical, impact-oriented orientation toward energy applications.