Distributed AI Support Engineer

About the role

GRNET S.A. provides internet connectivity, e-infrastructure and advanced services to the Greek educational, academic and research community. In 2026 it is expected to host the DAEDALUS supercomputer, based on NVIDIA GH200 architecture, which will also serve the Greek AI factory Pharos. As a Distributed AI Support Engineer you will help researchers, startups and industry teams turn this infrastructure into real-world AI breakthroughs, working with European universities, supercomputing centres and industrial partners in the EuroHPC ecosystem.

Key responsibilities

• First-line support for AI on HPC workloads (LLMs, computer vision, GPU-accelerated tasks): ticket triage, quick diagnosis of failed runs, escalation of suspected hardware issues.
• Help users write, review and debug Slurm job scripts and Ray/DeepSpeed and vLLM inference workflows.
• Maintain and test shared AI/LLM and computer-vision stacks (PyTorch, Hugging Face Transformers, DeepSpeed, vLLM, TensorFlow, RAPIDS, etc.) and design, publish and support Apptainer/Singularity containers.
• Diagnose failures (CUDA errors, NCCL timeouts, GPU OOM, distributed hangs) and profile/tune workloads with PyTorch Profiler, Nsight, TensorBoard, MLflow and Weights & Biases.
• Guide users on distributed training (DDP/FSDP, DeepSpeed ZeRO/pipeline/tensor parallelism), quantisation (QLoRA, 8-bit/4-bit) and large-scale inference (vLLM, TensorRT-LLM, Triton).
• Advise on storage layout, sharding and cleanup, and troubleshoot dataloader/I/O bottlenecks.
• Monitor usage metrics, support Access Call evaluations, and build documentation, tutorials and demos.
• Prepare technical reports, dashboards and web content, and coordinate announcements.

Required profile

• Degree in Computer Science, Engineering or a related STEM field; graduates and recent graduates considered.
• Strong Python programming skills and experience with AI frameworks.
• Hands-on experience training or fine-tuning models on GPUs with PyTorch and related tooling.
• Ability to explain technical concepts clearly to researchers and industry users, in writing and in person.

Required skills

• PyTorch, DDP/FSDP, Hugging Face Transformers, Accelerate, DeepSpeed, vLLM, Ray, TensorFlow, CUDA, NCCL, Slurm
• Apptainer/Singularity containers
• Desirable: GPU architecture knowledge, LLM training/fine-tuning, profiling tools, containerised GPU environments, HPC user support experience

What we offer

• Creative, dynamic and challenging working environment with opportunities for international collaboration
• Competitive remuneration package
• Opportunities for professional development