Postdoctoral Researcher in Confidential High Performance Computing for AI in Cancer Care (100%)

Your assignments

You will lead research on trustworthy and aligned LLMs operating in confidential HPC environments, involving training and adapting 70B-scale multilingual and medical language models on secure HPC infrastructure. Your work focuses on three areas:

• Preference optimization and LLM alignment: design preference-based training and fine-tuning methods (RLHF, PPO, DPO, reward modeling) for medical and multilingual LLMs.
• Agentic and tool-augmented AI systems: develop reasoning and interaction capabilities including RAG, in-context learning, chain-of-thought reasoning, and agentic workflows with external tools.
• Trustworthy AI evaluation and confidential deployment: develop evaluation pipelines (e.g., LLM-as-a-judge), robustness/bias/hallucination mitigation, and multilingual evaluation protocols.

As part of this position, your tasks also include: 
•    Conduct high-quality research aligned with the project objectives
•    Publish in top venues (NeurIPS, ICML, ICLR, ACL/EMNLP, MLSys, SC, HPDC, EuroSys, ASPLOS, etc.)
•    Present research at international conferences, workshops, and seminars
•    Contribute to open-source software and reproducible research artifacts
•    Collaborate actively, fruitfully, and respectfully with the PIs, the HPC group, and partner institutions
•    Contribute to teaching (max. one class per semester) and co-supervise Bachelor's/Master's students

Your profile

We seek a highly motivated researcher interested in integrating foundational AI and HPC research with real-world healthcare applications.

Required
•    PhD in Computer Science / AI / Machine Learning
•    Strong publication record in AI, ML systems, or related areas
•    Strong programming skills in Python, C/C++ and experience with PyTorch, TensorFlow, JAX, or similar ML frameworks
•    Experience with large-scale training or inference of LLMs
•    Interest in LLM alignment, reinforcement learning, or generative AI systems
•    Fluency in English; clear communication, problem-solving, and collaborative mindset

Highly desirable
•    Experience with distributed training or ML systems
•    Knowledge of privacy-enhancing technologies and parallel programming
•    Experience with multilingual AI or agentic systems

Diversity: We welcome applications from candidates with diverse backgrounds. Even if you do not meet all requirements, we encourage you to apply if you identify with the profile.

We offer you

•    Competitive 100% funding per SNSF guidelines (~CHF 90'000/year)
•    Access to modern GPU clusters and confidential-computing infrastructure
•    Collaboration with leading researchers in AI & HPC systems and digital health
•    Support for international conference travel and networking
•    Stimulating research environment at a leading European university