NeoBIM is a well-funded software start-up rethinking how architects design buildings through intelligent BIM tooling. As part of our applied research agenda, we are opening a thesis-aligned research position at the intersection of generative AI and computational architectural design.
Research Area
The broad focus of this position is the application of Generative Flow Networks (GFlowNets) to architectural design problems — in particular, the generation of diverse, constraint-aware spatial layouts. The concrete research question and scope will be developed together with the candidate based on their background and academic requirements.
Tasks
* Review relevant literature on GFlowNets and generative approaches to architectural layout generation
* Define and refine a research question in collaboration with the team
* Implement and experiment with GFlowNet-based models on spatial design tasks
* Design reward functions encoding architectural constraints and objectives
* Evaluate outputs in terms of diversity, feasibility, and constraint compliance
* Compare against suitable baseline generative approaches
* Document findings as a research report or thesis
Requirements
Must have
* Strong Python and PyTorch skills, as well as GIT
* Background in machine learning; familiarity with RL or probabilistic generative models is a plus
* Ability to work independently on open-ended research tasks
* Comfortable reading and engaging with recent ML papers
Nice to have
* Prior exposure to GFlowNets (Bengio et al.)
* Interest in architectural design or spatial reasoning
* Experience with graph neural networks or geometric deep learning
* Familiarity with graph-based data representations or knowledge graphs
Benefits
* A concrete, novel research problem with real-world grounding in AEC
* Close collaboration with our AI and product team
* Dedicated thesis mentorship from our AI team
* Potential for publication or continued collaboration post-thesis
* Flexible remote setup
For any questions regarding this position, feel free to contact Felix directly via phone or messenger at +49 176 95422094.