Challenges

Build a sentiment classifier for Qatari and Gulf Arabic social-media posts about government services, robust to code-switching and dialectal spelling variation. The winning model was hardened through the RAI pipeline and published as citizen-feedback-classifier, now serving MCIT's national feedback programme.

Forecast 15-minute vehicle counts across 120 Doha corridor sensors during Ramadan, when iftar and taraweeh travel patterns invert the normal peak structure. Models must generalise from three prior Ramadan seasons of loop-detector and event-calendar data supplied by MOTC.

Predict outpatient appointment no-shows across PHCC health centres using de-identified booking, weather, and historical attendance features. Winning models will feed an SMS nudge and overbooking optimiser expected to recover thousands of lost clinic slots per month.

Segment high-resolution satellite imagery of Qatar into fourteen land-use classes, from urban and industrial zones to sabkha flats and mangrove coastline. Outputs will refresh the national geospatial baseline used for planning, environmental monitoring, and coastal-resilience studies.

Transcribe conversational Gulf Arabic from government contact-centre calls, spanning Qatari, Bahraini, and Emirati accents with heavy English code-switching. Sadeem AI's winning system achieved 11.4 WER and was published to the marketplace as khaliji-asr, now powering voice services across three ministries.