Finalists work with regional air-quality and transport datasets
The main challenge used messy, realistic datasets with missing sensors, uneven city coverage, and time-based validation requirements.
A practical data challenge
The ML challenge asked finalists to model regional air-quality changes using weather, station, and transport signals. The dataset deliberately included gaps and inconsistent coverage so competitors had to reason about data quality.
Participants were expected to build a model, document feature choices, and avoid leakage across time windows. The judging team checked whether high scores were supported by a defensible validation plan.
Applied AI under constraints
The applied task connected the same skills to a public-service scenario. Competitors had to explain how their system would be used, when a human reviewer should intervene, and what failure cases mattered most.