Softquantus: Scaling SynapseX LLM for Quantum-Enhanced Financial and Industrial AI

SoftQuantus aims to scale its flagship model, SynapseX LLM, leveraging the LUMI-G EuroHPC JU infrastructure to advance sustainable, quantum-enhanced generative AI for industrial and financial domains. Over a 12-month period, the project will expand SynapseX from 7B to 13B–70B parameters, optimising its performance through federated incremental learning and Neighborhood Quantum Superposition for quantum-inspired optimisation.

The initiative directly addresses a major limitation of current Large Language Models — model drift — by enabling retraining on new data without reprocessing entire datasets, preserving prior knowledge and reducing energy consumption by 100–1000×. SoftQuantus will validate this approach on two key use cases: financial fraud detection and industrial predictive maintenance, achieving sub-100 ms latency in generative inference.

The project aligns with the EU’s Trustworthy and Sustainable AI goals, ensuring GDPR compliance, transparency, and bias mitigation. Results will demonstrate Europe’s capability to develop efficient, ethical AI at scale, bridging classical and quantum computing paradigms while strengthening digital sovereignty through resource-efficient innovation on EuroHPC JU systems.