DPFEED Reaches the Finish Line: A Real-Time Sports Data and AI Risk Platform, Built in Cyprus
We are proud to announce the successful completion of DPFEED (Datapulse Real-Time Feed Distribution and Risk Management), a nine-month research and innovation project delivered by Silicon Dali. Over the project period we took the platform from an early prototype to a fully qualified, production-ready service, hitting every major milestone along the way.
DPFEED was developed under the Cyprus Research and Innovation Foundation (RIF) RESTART 2016–2020 Fast Track Innovation programme, with co-funding from the European Union through NextGenerationEU and the Recovery and Resilience Facility.
What DPFEED Does
Sportsbook operators work in one of the most latency-sensitive environments in software. They rely on multiple external data providers, each delivering odds and event updates in slightly different formats and at different speeds, while simultaneously managing financial exposure across thousands of live markets. Doing this manually, across siloed systems, is slow and risky.
DPFEED closes that gap. It is a cloud-native, multi-tenant SaaS platform that ingests feeds from several providers in real time, normalizes and merges them into a single clean stream, applies intelligent risk and anomaly logic, and distributes a fast, reliable output through modern APIs and a management dashboard. In short, it brings feed distribution and risk evaluation together inside one unified system.
An AI Assistant for Split-Second Decisions
A standout outcome of the project is the platform's AI-assisted risk subsystem. Most betting slips clear automatically, but a risky minority are routed to a human trader who has only seconds to accept, decline, or re-price each one. DPFEED supports those traders with an AI layer that produces clear, explainable recommendations in real time.
The subsystem was built around a flexible dual-provider design, combining a hosted reasoning model with a self-hosted, domain-tuned model, all behind a single, consistent pricing pipeline. A key engineering breakthrough was a reasoning-first output format that lifted the compact model's decision accuracy to over 99 percent, matching full-precision performance while keeping the model small enough to run efficiently and affordably.
Milestones We Reached
The project advanced through six structured phases, and the team delivered on each one:
- Foundations validated. Confirmed the technical baseline and locked in a complete requirements and test plan with full performance, integration, and risk coverage.
- A faster core pipeline. Tuned the ingestion, messaging, caching, and database layers to sustain high-volume traffic with end-to-end latency consistently under 500 milliseconds.
- Intelligent risk built in. Integrated anomaly detection and risk management directly into the live data stream, validated under realistic peak-traffic conditions.
- A hardened, user-ready platform. Finalized the management dashboard, REST and WebSocket interfaces, and completed dedicated security audits.
- A successful local pilot. Deployed the full system in a live pilot in Cyprus, demonstrating stable performance and reliable risk processing under real usage.
- International validation and final qualification. Extended the platform to an international trial, completed formal performance, reliability, and security testing, and achieved full production-ready qualification.
Proven in Live Conditions
DPFEED was not just tested in the lab. It was validated in real pilot deployments in both Cyprus and an international market, where it maintained platform availability above 99.6 percent while sustaining thousands of messages per second across multiple concurrent data streams. The result is a service ready for real-world commercial use, backed by complete documentation, client onboarding materials, and training workshops.
Highlights at a Glance
- A full journey from early prototype to production-ready SaaS, delivered in nine months.
- Real-time feed aggregation, normalization, and merging across multiple providers.
- Sub-second, explainable AI recommendations for high-risk decisions.
- Compact AI model reaching over 99 percent decision accuracy.
- Above 99.6 percent availability across live pilots in two countries.
- Built on a modern, scalable cloud-native stack designed for growth.
Looking Ahead
With the platform now qualified for production, our focus turns to commercial rollout and continued innovation. We will keep advancing the AI subsystem, broaden integrations, and grow the platform's reach across new markets. DPFEED also strengthens Cyprus's position as an emerging hub for sports-technology innovation, and we are excited about what comes next.
Acknowledgements
DPFEED was made possible through the support of the Cyprus Research and Innovation Foundation under the RESTART 2016–2020 Fast Track Innovation programme, co-funded by the European Union (NextGenerationEU and the Recovery and Resilience Facility). We thank our pilot partners and the entire project team for their dedication in bringing this platform to life.
Silicon Dali develops cloud-native software for the sports-technology sector, with a focus on real-time data and intelligent risk management.
