At SpiNNcloud, I contributed runtime and infrastructure software for a neuromorphic supercomputer with approximately 5 million cores. I drove the architecture and implementation of three hardware allocators, including topology-aware placement for a Slurm-managed system.
I developed deterministic allocation strategies for a 2D torus and contributed to graph discovery, reachability, routing, and the UDP-based SpiNNaker2 Datagram Protocol. I also developed Python job-management components and a C++ API for hardware requests and reservations.
At Rohde & Schwarz, I optimized a signal-transformation library to run 2.5x faster and debugged Linux, QEMU, PTP, and O-RAN issues. In my open-source project, I’m building a job-scheduling and topology-aware resource-management platform for simulated HPC/GPU clusters.

