CASTOR Horizon will be represented at the IEEE Computer Society Annual Symposium on VLSI (ISVLSI) 2026 by the University of Amsterdam. UvA Ph.D. student Hang Xu will present the paper “Fast Transformer Inference on ARM-Based HMPSoCs.”
UvA is primarily responsible for developing and implementing optimization algorithms in the optimization engine in CASTOR. In this context, the presented research focuses on optimisation algorithms for neural network inference on edge devices. The graph optimisation expertise developed through this work is also relevant to trusted path routing across computer network graphs, contributing to CASTOR’s objectives in secure and efficient network orchestration.
The paper was developed in collaboration with fellow CASTOR partner UBITECH, further strengthening the research cooperation within the consortium.
