What matters
The problem comes before the model.
- Point-in-time vintages and prevention of look-ahead bias.
- Comparison with simple baselines before assigning value to more complex models.
- Intervals, probabilities and prospective error tracking rather than retrospective charts alone.
Lab Intelligence approach
Sources, assumptions, checks, versions, metrics and limitations remain visible. A more complex technique is used only when it adds value over an understandable baseline.
How the method works →Related workflows
Examples already documented.
These pages describe implementations or experiments from the laboratory; they do not imply that the same result transfers automatically to another context.
World Economy Engine
Integrates macro and market data into a Global Economy Pulse, time-causal regimes, a structural graph, scenarios and forecasts with prequential selection, benchmarks and uncertainty intervals.
Open project →WF-2The New Margin
Measures the 3-2-1 crack spread, produces multi-horizon forecasts and preserves a mature-error track record.
Open project →WF-3Market Overview
Condenses cross-asset signals into regime, forecast, anomalies and a risk-on/risk-off view while keeping uncertainty visible.
Open project →WF-13Italy Variable Renewable Forecast
Forecasts the Europe/Rome D+1 civil day for solar, wind and total VRE, normalising by capacity and archiving point-in-time vintages.
Open project →WF-7Etna Forecast
24-hour and 5-day probabilistic nowcasting with public multimodal data, strong baselines, a modality gate and declared degradation.
Open project →WF-17Etna Fusion Lab
Fuses tremor, seismicity, thermal signals and event history with quality-adjusted weights; separates the 1h/6h Activity Nowcast from the 1d/5d VONA ash-positive forecast and keeps a prospectively verifiable ledger.
Open project →WF-1Weather automation
Turns public weather forecasts into an educational, readable and quality-controlled multi-city index.
Open project →WF-20The Public Service Run
Educational workflow that compares traffic-aware routes and departure windows for public-service commuting scenarios, combining routing, weather, historical calibration, Random Forest after warm-up and P10/P50/P90 uncertainty.
Open project →Want to discuss a similar case?
Describe the problem and expected result in non-confidential terms. The first step is to identify the data, criteria and limitations that would make the case verifiable.