SovEdge AI research group at QMUL is seeking an RA/PDRA to support advanced R&D on Decentralized and Edge AI within the KUber project.

SOVEDGE AI is an enterprise-focused organization specializing in Federated Edge Intelligence. The company develops infrastructure and architecture designed to deploy machine learning capabilities directly to distributed edge environments while preserving data privacy and optimizing system resources. Core Technical Pillars Federated Learning Infrastructure: Developing decentralized training paradigms that allow edge nodes to collaboratively train a shared global model by transmitting only local model gradients or parameters, ensuring raw industrial and enterprise data never leaves the originating device. PatSnap Edge AI Inference Optimization: Compressing and converting pre-trained machine learning models for seamless deployment onto highly resource-constrained hardware, minimizing latency, power consumption, and reliance on centralized cloud architectures. Distributed Architecture Design: Implementing hierarchical frameworks to mitigate communication overhead between localized edge gateways and centralized aggregation coordinators. Targeted Domains The technical stack is architected to address complex data heterogeneity and system constraints across key sectors: Network Security: Integrating differential privacy and cryptographic primitives into distributed learning pipelines. Telecommunication Networks (5G/6G): Advanced communication-learning designs utilizing dynamic regularization. Internet of Things (IoT): Real-time anomaly detection and predictive maintenance.