- Describe your company in one paragraph.
Multiverse Computing is the leading compressed AI model provider. The company’s deep expertise in quantum software and AI led to the development of CompactifAI, a revolutionary AI model compressor. CompactifAI compresses LLMs by up to 95% with only 2-3% precision loss. The company now offers compressed AI models that reduce computing requirements and unleash new use cases for AI across industries.
Multiverse Computing is headquartered in Donostia, Spain, with offices across Europe, the U.S., and Canada. For more information, visit multiversecomputing.com
- What challenge are you addressing under the O-CEI Horizon’s first Open Call, and how is your proposal relevant to the challenge?
We are addressing Challenge P3C4 (Pilot 3), tackling the critical bottleneck in deploying AI for smart electric vehicle (EV) fleet management: the lack of rich, granular, and shareable time-series data. Optimizing depot charging, Vehicle-to-Grid (V2G) participation, and PV-coupled energy systems requires vast amounts of operational data. However, real-world data is severely restricted by privacy regulations (GDPR), fragmented across subsystems, and critically lacks rare but high-risk “edge cases” (e.g., simultaneous fast-charging peaks or sudden solar drops). Standard, brute-force generative AI fails here because it merely mimics statistical patterns, often producing physically impossible energy scenarios.
Our solution, SynthiaFleet, solves this by delivering a constraint-aware, privacy-preserving synthetic data generator. We differentiate our approach by leveraging our core expertise in physics-based AI and Tensor Networks. Instead of relying on blind statistical replication, we embed actual physical constraints—State of Charge (SoC) dynamics, depot power limits, and weather-dependent PV curves—directly into the generative process. Furthermore, we integrate strict Differential Privacy to ensure the synthetic datasets are entirely decoupled from real, identifiable fleet behaviors, enabling safe, federated data sharing across the Cloud-Edge-IoT continuum. Finally, SynthiaFleet’s edge-case module systematically extrapolates extreme, under-represented operational stress conditions, ensuring that downstream algorithms are trained for real-world reliability, not just average-case scenarios.
- What is the expected impact of your proposal?
SynthiaFleet delivers a dual impact: enabling a faster, safer transition to electrified mobility while promoting responsible, Trustworthy AI under the EU AI Act.
Commercially, we are establishing a highly scalable “Generator-as-a-Service” for the O-CEI Marketplace and the broader European market. By providing physically plausible synthetic data, we remove the primary roadblock for grid operators, logistics fleets, and digital-twin vendors who need to train advanced algorithms but are blocked by data scarcity. This accelerates the deployment of AI-driven energy management systems across Europe.
Environmentally, SynthiaFleet acts as a catalyst for Green AI. Directly, it eliminates the need for energy-intensive, physical load testing by providing high-fidelity digital testing environments. Indirectly, the optimization algorithms trained on our synthetic data will allow fleets to shift demand to low-carbon hours, maximize the utilization of local solar generation, and enable reliable Vehicle-to-Grid (V2G) participation. This prevents unnecessary battery degradation and reduces reliance on fossil-heavy grid imports.
Socially, SynthiaFleet strengthens data sovereignty and societal trust. By embedding Differential Privacy into our generative mechanics, we ensure sensitive corporate and behavioral data is never exposed. We allow innovators to build high-stakes mobility solutions without compromising individual privacy. Through SynthiaFleet, we prove that the European AI ecosystem does not need to rely on brute-force data harvesting; instead, we can use smart, physics-based generation to achieve superior, privacy-respecting results.
