Master’s Thesis: Engineering-Aware GenAI - Grounding and Validation for Reliable CAD Model Generation
Master's Thesis, Aktuelles |
Recent advances in generative AI enable the creation of text, images, and increasingly 3D objects. In engineering design, this opens up new possibilities for AI-assisted CAD model generation. Instead of manually modeling geometry, designers can collaborate with GenAI systems. However, a significant gap remains between general-purpose generative models and the requirements of professional engineering. For GenAI to be a reliable design tool, it must be grounded in domain-specific standards, and its outputs must be validated for mechanical viability.
Objective
The goal of this work is to develop and evaluate methods to ground GenAI in engineering reality, ensuring that co-creative CAD generation follows technical standards and produces viable results.
This may include:
- Knowledge Integration: Researching and implementing methods (e.g., Retrieval-Augmented Generation) to inject domain-specific knowledge, design guidelines, and part libraries into the AI workflow.
- Technical Validation: Developing "guardrails" or validation layers that automatically check AI-generated geometry against mechanical rules and quality metrics. This may include evaluating feasibility, part completeness, and detecting geometric defects.
- Feedback Loops: Designing interaction mechanisms that allow the system to critique its own output and self-correct based on engineering constraints.
- Evaluation: Testing the system's ability to adhere to engineering standards.
Requirements
- Enrolled in a Master's degree in Electrical Engineering, Computer Science, or Mechanical Engineering
- Strong ability to work independently
- Very good programming knowledge (e.g. Python)
- Solid knowledge of software development principles
- Proficient with Git
- Basic understanding of machine and deep learning
- Interest in the intersection of AI and Computer-Aided Design
To apply, please write an email to david.fresacher@tum.de with a short motivation, an overview of your relevant skills and experience, and your ideas for a specific focus within this scope.