AI for Science Engineer

About the role

The ETH AI Center, one of the largest AI centers in Europe, is looking for an AI for Science Engineer in Zurich for a fixed-term position covering 80% to 100% of full time. You will build frontier AI systems that help researchers across ETH accelerate their research, partnering with research labs to co-develop agentic solutions such as multi-step pipelines that plan and orchestrate simulations and experiments, curate data, evaluate models or chain domain-specific tools.

Key responsibilities

• Partner with research labs to identify high-impact use cases for agentic AI and co-develop novel pipelines and tools.
• Build planning and orchestration components that turn research goals into executable workflows under constraints such as cost, latency and fidelity.
• Make system behaviour inspectable: traceable runs, reproducible results and interfaces that let researchers understand and steer the work.
• Support research teams with data preparation and curation, and help them optimise, specialise and deploy models (fine-tuning, distillation, quantisation, efficient inference).
• Contribute to evaluating and benchmarking AI systems on scientific tasks, including frontier models.
• Develop and open-source reusable recipes, reference architectures and templates.
• Evaluate emerging frameworks, prototype new patterns and advise research teams.
• Contribute to teaching and student projects.

Required profile

• Doctorate in computer science or a computational discipline preferred; a master's degree with strong relevant experience is also welcome.
• 1-3 years of experience building and deploying agentic or LLM-based systems end to end; recent graduates are encouraged to apply.
• Ability to engage with researchers on their own terms.
• Fluency in English.

Required skills

• Services, APIs, data pipelines, containers, CI/CD.
• LLM APIs, agent frameworks, RAG pipelines, PyTorch, HuggingFace, model optimisation and efficient inference.
• GPU computing and high-performance or distributed computing environments.
• Planning, scheduling, workflow engines or reasoning systems, and scientific computing workflows are an advantage; open-source contributions are a strong plus.