Summary
Manos Pavlidakis is a software engineer and researcher with nine years of experience focused on secure, high-performance sharing of heterogeneous accelerators (GPUs, FPGAs, TPUs) for HPC and ML workloads. He completed a PhD-driven research program at FORTH where he designed practical systems like G-Safe, Arax, and TReM that blend PTX/CUDA/ROCm internals, IPC, and runtime scheduling to enable safe, efficient multi-tenant GPU use. Currently at Spectral Compute, he helps advance the SCALE compiler and language to make CUDA workloads portable to AMD GPUs, bridging major vendor gaps in GPGPU programming. He holds two NVIDIA certifications for scaling CUDA across nodes and combines deep low-level expertise (binary instrumentation, LD_PRELOAD, CUDA PTX) with runtime and compiler work. Notably, his background spans both FPGA tooling (SDAccel/Vinetalk) and distributed accelerator orchestration, giving him a rare end-to-end perspective on accelerator ecosystems. Based in Heraklion, Greece, he is driven to translate research prototypes into production-ready, cross-platform accelerator solutions.
9 years of coding experience
3 years of employment as a software developer
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at University of Crete, Department of Computer Science
Master's degree, Computer Science, 7.6, Master's degree, Computer Science, 7.6 at Vrije Universiteit Amsterdam (VU Amsterdam)
BSc, Computer and information systems, BSc, Computer and information systems at Panepistimion Aegaeou
English