Projects

From papers to products.

MacroLab AI projects are grounded in research and delivered as working applications. Below is what we are building, and the research behind it.

Projects

Reading the business cycle in real time.

Real-time reads on the state of the economy, built from multimodal data-streams. Our models process thousands of articles as they are published and combines them with hard economic indicators to identify business-cycle fluctuations and the narrative that moves it.

Because the story matters, filtering is not without structure and can be decomposed: every movement in the cycle can be traced to demand and supply inovations, and ultimately to the individual articles behind it.

All of this comes together in a live dashboard. When the indicator moves, it shows which narratives moved it and which articles carried them, so a surprise in the data always arrives with an explanation.

Related research:

Open dashboard Coming Soon
Projects

Forecasts, scenarios and what happens if we are wrong.

Forecasts that use more of the information that is actually available: published projections, prediction records, and announced policy paths, read together with the realised data.

The same tools support risk management. Automated research workflows run scenarios and estimations in parallel, so we can ask both what we expect and what happens if we are wrong.

Announced policy paths get particular attention. Our research shows how to read conditional forecasts directly out of central bank communication, so the announced path becomes a testable input to the forecast.

Related research:

Open dashboard Coming Soon
Projects

Measuring what the economy talks about.

Quantitative indexes built from decades of text: sentiment, economic uncertainty, and the themes that drive them. Machine-learning classification separates types of news that keyword methods lump together.

The result is a set of series that can be followed over time, compared across topics, and used directly in models, with crisis episodes and turning points visible in the data.

Because every index is built the same way, new ones are cheap to add: from broad economic uncertainty down to narrow series like climate risk in commodity markets. And every series runs back decades, so today’s reading can be set against every episode before it.

Related research:

Open dashboard Coming Soon
MacroLab AI · Economic analysis with artificial intelligence