Thesis Internship: Photo‑Realistic Virtual Training Data for Physical AI

About the role

This thesis internship focuses on generating photo‑realistic virtual training data for physical AI models used in robotics. You will work within ABB's Robotics business, collaborating with R&D experts to explore simulation fidelity and AI performance.

Key responsibilities

• Research and evaluate simulation methods for capturing large‑scale robot video footage.
• Create 3D digital twins of robot setups using techniques such as Gaussian Splatting, Triangle Splatting, and mesh generation.
• Assess physical AI models (e.g., ACT – Action Chunking with Transformers) trained on virtually generated video data.
• Determine the optimal balance between simulation speed, complexity, and AI performance.

Required profile

• Engineering student in robotics, computer science, or a related field.
• Experience with machine learning, AI vision systems, or imitation learning.
• Strong knowledge of linear algebra.
• Hands‑on experience with ABB industrial robots is a plus.

Required skills

• Robotics simulation and 3D rendering.
• Gaussian Splatting / Triangle Splatting.
• Mesh generation.
• Machine learning and AI model training.
• Action Chunking with Transformers (ACT).
• CAD or 3D modelling techniques.