Large-Scale AI Engineering
Hands-on ETH Zurich course led by Imanol Schlag, which I helped initiate and co-lecture. Students train and optimize large models on 32 GH200 GPUs each on CSCS's Alps supercomputer.
Large-Scale AI Engineering is an ETH Zurich course led by Imanol Schlag (Technical Lead for LLM Development, Swiss AI Initiative), first offered in Spring 2025. I helped initiate the course and am one of its lecturers; Imanol designed and builds out its content. It lives at the intersection of high-performance computing and generative AI: while many universities teach HPC or machine learning, few cover the engineering skills actually needed to train and optimize large-scale AI systems.
What students do. Beyond lectures, students complete six assignments and a final mini-project on distributed training of large language models, and are encouraged to merge their code with other students’ features — the collaborative workflow they will meet in practice. Each team gets access to 8 nodes (32 NVIDIA GH200 Grace Hopper superchips) on Alps, the supercomputer at the Swiss National Supercomputing Centre (CSCS) inaugurated in fall 2024 — the largest AI-ready system hosted at a public institution. CSCS engineers also join the course to present the center’s mission and infrastructure.
Reach. Although a Master’s course (3 ECTS), roughly 25 attendees of the first edition were PhD students or postdocs, and the Department of Computer Science has since expanded the number of seats. The course has run every semester since: Spring 2025, Fall 2025, Spring 2026, and Fall 2026.