A shift in wind or unexpected rain can shift power markets in minutes. Jua's proprietary data models captured these shifts early. The challenge was interpreting this data fast enough.

Intraday markets are highly-volatile, and operate in narrow 15-minute intervals. Resulting in positions being constantly adjusted.

Traders only act when accuracy of data and speed converge. If a trader adjusts their position too late, they can lose millions — if not tens of millions of dollars.

The goal was to shorten the distance between identifying forecast change, and responding to market movement. Every surface had to justify its place within the product.

An AI-generated morning briefing summarises the forecast. It helps traders understand market conditions without needing to parse raw forecasting data.

The map gave traders fast geospatial context. This lets a trader see where changes to the forecast intersect with regions and trading zones.

Scenarios frame a set of plausible market realities. Each scenario carries a confidence score that helps traders understand market risk.

The activity feed surfaced forecast changes as they happened throughout the day. It removed the need for traders to manually re-run models and drive up processing costs.

A conversational agent let traders ask questions and get clear answers about the forecast. It lowered the technical floor required to access proprietary forecasting models.

I rapidly designed and engineered the product inside Cursor. Using various models, such as Sonnet 4.5, and GPT 5.1 Codex. I then connected Open-Meteo's API to return live forecasting data through the agent.

Product direction was reviewed with operations, engineering and stakeholders. Feedback centred around users being able to interpret and act on the weather intelligence quickly.

The final product allows traders to make faster and more informed decisions. Making dense and complicated data easy to access — and act quickly upon.