Integrating CAD and AI: Autoregressive and Diffusion Models for Intelligent Design
STC on “Computer Aided Design and Artificial Intelligence” – NITTTR Chandigarh
As part of the Short Term Course (STC) on “Computer Aided Design and Artificial Intelligence”, organized by the National Institute of Technical Teachers Training and Research (NITTTR), Chandigarh, and conducted from 01 to 05 December 2025, two sessions were delivered on 04 December 2025. The sessions focused on how modern generative AI models are transforming engineering design, CAD workflows, and visualization practices.
The first session, held from 10:00 AM to 11:30 AM, focused on Autoregressive Models for AI-Driven Design and Code Generation. Participants were introduced to the fundamentals of autoregressive modeling, where outputs are generated sequentially based on prior context. The session explained how such models form the backbone of large language models and are increasingly used for generating structured engineering outputs such as parametric CAD scripts, design logic, and technical documentation.
Through practical examples, the session demonstrated how natural language prompts can be translated into executable CAD and geometry scripts using tools like FreeCAD and OpenSC. Emphasis was placed on understanding design intent, prompt structuring, automation of repetitive modeling tasks, and the role of autoregressive models in parametric and rule-based design systems. Limitations related to geometric constraints and manufacturability were also discussed to provide a balanced perspective.
The second session, conducted from 3:00 PM to 4:30 PM, focused on Diffusion Models for Generative Design and Visualization. Participants gained conceptual clarity on how diffusion models generate designs by progressively refining noise into structured outputs. The session highlighted why diffusion models are particularly effective for creative design exploration, early-stage ideation, and high-quality visual generation, rather than strict parametric modeling.
Through hands-on demonstrations and prompt examples, participants explored how diffusion-based tools such as Stable Diffusion and related open-source platforms can be applied to product design, architectural visualization, UI/UX layout exploration, and early-stage CAD inspiration. Real-world use cases showed how diffusion models help designers and engineers visualize forms, textures, and structures before detailed CAD modeling, thereby accelerating creativity and reducing iteration time. Best practices and limitations were also discussed, including the need for constraints, human-in-the-loop refinement, and integration with structured CAD systems.
Together, the two sessions offered a complementary perspective on generative AI for engineering design. While autoregressive models enable structured, rule-based generation of CAD and engineering code, diffusion models support creative exploration and visualization. This combined understanding equips educators and researchers to bridge traditional CAD methodologies with modern AI-driven design paradigms, aligning well with emerging trends in intelligent manufacturing, generative design, and Industry 4.0.
Gratitude is extended to NITTTR Chandigarh for organizing this forward-looking ICT-based STC. Special thanks to Dr. Amit Doegar, Associate Professor, Department of Computer Science & Engineering, and Dr. P. Sudhakar Rao, Associate Professor, Mechanical Engineering Department for their kind invitation and coordination of the program. Appreciation is also extended to the faculty and all participants for their enthusiastic engagement, curiosity, and insightful questions, which made the session highly interactive and enriching.