TY  - JOUR
AB  - Mechanical Computer-Aided Design (CAD) remains a challenge in product development and engineering design due to steep learning curves. Hence, large language models (LLM) gain popularity to generate parametric geometry programmatically (CAD-as-Code). This paper examines CAD-as-Code as a methodological paradigm with implications for engineering design practice. We present a hybrid agentic framework, that uses CAD-as-Code with physics-in-the-loop and human validation to generate structurally sound designs. We prove our concept by generating physically valid designs based on load bearing requirements. Finally, we derive implications for engineering design practice from our prototype observations.
AU  - Berger, Elias
AU  - Mehlstäubl, Jan
AU  - Saske, Bernd
AU  - Paetzhold-Byhain, Kristin
DO  - 10.17619/UNIPB/1-2656
PB  - Universitätsbibliothek
DP  - Universität Paderborn
LA  - eng
PY  - 2026
SP  - 1 Online-Ressource (Seite 129-138) : Diagramme, Illustrationen
T2  - 1st International Symposium: March 24 – 26, 2026, Heinz Nixdorf Institute, Paderborn University
TI  - CAD-as-Code and agentic AI for function-aware engineering design
UR  - https://nbn-resolving.org/urn:nbn:de:hbz:466:2-58933
Y2  - 2026-09-18T20:41:07
ER  -