Filippo Ziche

Filippo Ziche is a PhD student at the University of Verona and a member of PARCO. His research studies how parallel and heterogeneous computing can make data-intensive algorithms practical on GPUs and resource-constrained edge platforms. His work on dynamic graphs includes GPU-accelerated breadth-first search that updates traversal results efficiently as a network changes, rather than recomputing them from scratch.

A second line of his research concerns efficient human-action analysis at the edge. He co-developed OLORAS, a temporal action-segmentation model that uses human-pose sequences and linear recurrent units to process long activities with low memory requirements. This connects PARCO’s work on high-performance and edge computing with applications in healthcare and Industry 5.0, where latency, privacy and limited on-device resources constrain how perception models can be deployed.