Horizon captures real Saudi driving and turns it into PDPL-compliant datasets and benchmarks, so the Kingdom can decide which self-driving systems are safe to deploy on its roads.
Real Saudi driving captured with Comma 4X kits: camera, GNSS, IMU, and CAN, geo-tagged across multiple cities and turned into clean, compliant sequence data.
Automated deduplication, semantic indexing, and edge-case and long-tail retrieval surface rare scenarios. Model-assisted annotation with human QA, plus full versioning and lineage.
We don't just deliver data, we prove it. Horizon trains and evaluates AV models on Saudi data and ships ODD benchmarks and gap reports that show where foreign-trained models fail.
Foreign datasets miss Saudi reality. Our data encodes bilingual signage, local infrastructure and roundabouts, a distinct vehicle mix, aggressive merges, and desert conditions: dust, glare, and heat.
Horizon enforces Saudi data-sovereignty, privacy, and security at the platform level, aligned with PDPL, SDAIA, NCA, and GEOSA: every byte stays within the Kingdom's regulatory perimeter.
Captured and processed by GEOSA-licensed Saudi entities on sovereign compute, encrypted, and stored entirely in the Kingdom under SDAIA and PDPL. Zero cross-border transfer.
PDPL-compliant redaction anonymizes faces, plates, and identifiers. Role-based access and audit trails apply, and residency is enforced, so nothing leaves Saudi Arabia.
Sensitive sites and assets are masked, with encryption and tamper-evident logging. Workflows align with NCA cybersecurity controls, SHC, and SASO AV regulations.
The first sovereign data and benchmark platform built for Saudi autonomy. Capture, governance, curation, annotation, training, and benchmarking, all in-Kingdom, with no foreign dependency.
Build the data foundation for safe AV validation and certification under Vision 2030, so teams can train and benchmark on data that reflects real Saudi roads.
Models trained on foreign data fail on Saudi roads. Horizon closes the gap by mining Saudi-specific edge cases: heat, dust, glare, bilingual signage, aggressive merges, and dense intersections.
Built for government programs, AV and ADAS developers, OEMs and Tier-1s, and mobility operators that need sovereign, Saudi-specific data and benchmarks.
Beyond datasets: AV readiness assessments, model and perception benchmarking, safety-validation audits, and SASO/SHC-aligned regulatory evaluations, from prototype to certification.