PostDoc Position – Physical AI Research

 

As a PostDoc researcher and co-founder of the Physical AI group, you will develop methods that enable robots to learn from
demonstrations, corrections, and autonomous experience, and deploy them in real-world settings.

  • Vision-Language-Action (VLA) Models: Design and implement VLA architectures integrating vision, language, and action for dexterous manipulation, building on large pre-trained vision-language backbones (e.g., 5B-parameter VLMs).
  • Reinforcement Learning from Experience: Develop RL pipelines – offline RL, advantage-conditioned policies – enabling robots to grow beyond pure imitation, achieving human-level and superhuman robustness through autonomous experience.
  • Long-Horizon Task Mastery: Investigate credit assignment across extended tasks via learned value functions, enabling robots to detect and correct compounding errors in complex real-world scenarios.
  • Sim-to-Real Transfer & World Models: Bridge simulation and deployment using world models, self-supervised representations (JEPA, DINOv3), and transfer techniques for robust generalization.
  • Medical & Clinical Robotics: Partner with imaging and clinical groups to apply Physical AI in healthcare robotics, combining LFB’s sensor expertise with embodied intelligence.

Candidate Profile:

  • PhD in Computer Science, Electrical Engineering, Robotics, Physics, or a related field
  • Strong background in deep learning, with experience in reinforcement learning, imitation learning
  • Hands-on experience with PyTorch and large-scale model training; familiarity with VLA or foundation model architectures is a strong advantage
  • Publication record in top-tier venues (NeurIPS, ICML, ICLR, CoRL, ICRA, or equivalent)
  • Drive to work at the intersection of Physical AI, embodied intelligence, and real-world deployment
  • Excellent communication skills in English; German is advantageous but not required

 

PhD Position: AI for Automated Surgical Planning

 

Your Research Impact
As a PhD researcher, you will develop AI methods that automate surgical planning.

  • 3D Anatomical Intelligence
    Develop representation learning approaches based on DINOv2 for high-resolution CT and CBCT/DVT data to
    capture complex anatomical structures.
  • Self-Supervised Learning
    Design learning strategies that leverage large volumes of unlabeled medical imaging data.
  • Multimodal Surgical Reasoning
    Combine visual models with multimodal language models to translate clinical descriptions into surgical
    planning instructions.
  • Automated Planning
    Integrate AI representations with geometric algorithms to propose osteotomy planes and surgical plans.

Candidate Profile:

  • Very good Master’s degree in Computer Science, Physics, Biomedical Engineering, or a related field
  • Strong background in AI/ML and experience with Python/PyTorch
  • Interest in self-supervised learning or vision transformer models
  • Motivation to work on interdisciplinary medical AI challenges

HiWi Offers

 

We continuously hire HiWis and would be happy to hear from you—feel free to come and talk to us anytime!.

 Studentische Hilfskraft (m/w/d) im Bereich Videosignalverarbeitung
Ansprechpartner: Mathias Wien