Runtime-Integrated Simulation of Accelerator Memory Behaviour for Distributed LLM Deployment

Master internship, PhD internship - Leuven | About a week ago Developing hardware for AI is both costly and technically demanding. The software stack evolves rapidly and is often optimized around GPUs. These realities raise a fundamental challenge: how can one validate hardware that does not yet exist for emerging AI workloads? Fortunately, most AI applications are built on a relatively small set of frameworks, notab…