CDFAM Computational Design Symposium

Duann Scott

CDFAM Computational Design Symposium Presentation Recordings www.designforam.com

  1. 19H AGO

    AI Judges In Design: Statistical Perspectives On Achieving Human Expert Equivalence With VLMs

    Recorded at CDFAM Computational Design Symposium, Amsterdam, 2025 https://cdfam.com/amsterdam-2025/ Organization: MIT Presenter: Kristen Edwards Presentation Abstract The subjective evaluation of early stage engineering designs, such as conceptual sketches, traditionally relies on human experts. However, expert evaluations are time-consuming, expensive, and sometimes inconsistent. Recent advances in vision-language models (VLMs) offer the potential to automate design assessments, but it is crucial to ensure that these AI “judges” perform on par with human experts. However, no existing framework assesses expert equivalence. This research introduces a rigorous statistical framework to determine whether an AI judge’s ratings match those of human experts. We propose statistical metrics that broadly cover these assessment areas: interrater reliability, agreement, error metrics, correlation and relative rank assessment, distribution-similarity analysis, and equivalence tests. We apply this framework in a case study evaluating four VLM-based judges on key design metrics (uniqueness, creativity, usefulness, and drawing quality). These AI judges employ various in-context learning (ICL) techniques, including uni- vs. multimodal prompts and inference-time reasoning. The same statistical framework is used to assess three trained novices for expert-equivalence. Results show that the top-performing AI judge, using text- and image-based ICL with reasoning, achieves expert-level agreement for uniqueness and drawing quality and outperforms or matches trained novices across all metrics. This has implications for scaling design evaluation in education and practice, and provides a general statistical framework for validating AI judges in other domains requiring subjective content evaluation. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.designforam.com

    17 min
  2. AUG 25

    Accelerating Design Optimization Using Implicit Geometry : Bradley Rothenberg + Max Gaedtke - nTop

    Recorded at CDFAM Computational Design Symposium, Amsterdam , 2025 https://cdfam.com/amsterdam-2025/ Organization: nTop Presenter: Bradley Rothenberg + Max Gaedtke Brad Rothenberg, CEO of nTop, outlines a shift toward co-intelligent engineering—a framework where human intent is augmented by high-performance modeling and simulation tools to meet the increasing demands of modern engineering timelines. The talk explores how traditional CAD and simulation workflows fall short under compressed development cycles, particularly in aerospace, and introduces a modular, parametric approach to design that supports rapid iteration, reuse, and performance-driven exploration. Rothenberg highlights how implicit modeling and GPU-native solvers can be combined to eliminate bottlenecks in geometry creation and simulation, enabling faster, more informed decision-making. The broader message emphasizes a move from digitizing drawings to capturing design intent—using tools that work withengineers to accelerate innovation without compromising performance or rigor. — Learn more about the CDFAM Computational Design Symposium series—including past presentations and upcoming events—at https://cdfam.com #CDFAM #nTop #ComputationalDesign #ImplicitModeling #EngineeringWorkflow #CoIntelligentEngineering #Simulation #DesignAutomation #AerospaceDesign #CFD #GenerativeDesign This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.designforam.com

    25 min
  3. AUG 21

    SimScale – Physics & AI engineering simulation in the cloud - David Heiny

    Recorded at CDFAM Computational Design Symposium, Amsterdam , 2025 https://cdfam.com/amsterdam-2025/ Organization SimScale Presenter: David Heiny CEO & Co-Founder In this presentation from CDFAM Amsterdam 2025, David Heiny, co-founder and CEO of SimScale, outlines a pragmatic approach to integrating AI into engineering workflows. Drawing from conversations with engineering leaders and data collected from a recent industry survey, he shares key insights into why widespread adoption of AI in core engineering remains limited—despite significant advances in generative design, simulation acceleration, and surrogate modeling. The talk explores: * The structural complexity of modern engineering organizations * Why conventional AI success in image generation or code translation hasn't translated to simulation workflows * A layered view of how AI tools are beginning to augment specific stages of the product development cycle Heiny introduces SimScale’s approach to cloud-native simulation, including the use of frontier models for automating setup and interpretation, and the deployment of physics-based AI surrogates for fast, scalable parametric optimization. A live demonstration showcases how AI agents can assist in simulation setup and design evaluation directly within SimScale's platform. Additional topics include: * AI adoption trends from a 300-leader survey * The role of data structuring and cloud infrastructure in enabling automation * Use cases from industry and academia, including work on snap-fit prediction, pump optimization, and native integration of implicit geometry via nTop This presentation highlights the emerging convergence of simulation, design data, and AI, and presents a compelling case for how engineering teams can begin implementing these capabilities today. — To learn more about the CDFAM Computational Design Symposium series, including previous presentations and future events visit https://cdfam.com #CDFAM #ComputationalDesign #AIinEngineering #Simulation #CAE #DesignAutomation #SimScale #nTop #PhysicsAI #EngineeringWorkflow #GenerativeDesign #SurrogateModeling This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.designforam.com

    22 min
  4. AUG 18

    Flexible Geometric Modeling and Atypical Simulation Solvers to Streamline Design Optimization - Wesley Essink - Altair

    Recorded at CDFAM Computational Design Symposium, Amsterdam, 2025 https://cdfam.com/amsterdam-2025/ Organization: Altair Presenter: Wesley Essink Flexible Geometric Modeling and Atypical Simulation Solvers to Streamline Design Optimization Presentation Abstract Simulation-driven design serves two important purposes: wider exploration of the design space and goal-seeking optimization. Regardless of the regime, the workflow spanning geometry creation, simulation setup, results interrogation and geometry redesign needs to be as seamless as possible to make this approach viable. However, there is often a significant overhead associated with manual, non-value-adding tasks. Examples include converting all geometry into a common representation prior to simulation, meshing for simulation, (re)applying simulation boundary conditions and, finally, making meaningful geometry updates based on simulation results. In this talk, we will showcase some of the approaches and methods we use in Altair Inspire to: Concurrently model with up to four different geometry representations Prepare simulation boundary conditions that remain fixed, irrespective of geometry changes Run simulations on components and assemblies without having to harmonise all geometry into a single representation Prepare a design exploration or optimization to close the loop between design and simulation Automatically update geometry based on the simulation findings These workflows take place entirely within Altair Inspire, which also reduces the need for lossy conversions or file transfers between different software products. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.designforam.com

    20 min
  5. AUG 14

    From Structure To Sound: Unlocking The Potential Of Vibroacoustic Design - Vanessa Cool - KU Leuven

    Organization: Ku Leuven Presenter: Vanessa Cool From Structure To Sound: Unlocking The Potential Of Vibroacoustic Design Presentation Abstract Noise pollution is one of the leading global environmental pollutants, resulting in the loss of one million life years annually. This has prompted the introduction of stringent regulations on the acoustic performance of structures, without compromising their structural integrity. Topology optimization offers a promising approach to developing innovative structures that meet these conflicting requirements. However, in many applications, considering vibroacoustic coupling from the early design stages is essential, as it directly impacts the accuracy of the acoustic performance and structural stability, ensuring that both functional and regulatory requirements are met. This added complexity to the optimization process largely influences the resulting structures and is crucial for achieving optimal performance. This presentation will provide a comprehensive overview of an intricate vibroacoustic topology optimization framework and focusses on its potential applications. It will showcase optimization results across various scales, from unit cell and metamaterial design to supercell and finite component levels. Novel, intricately engineered structures that balance lightweight design, structural stiffness, and acoustic performance will be presented, demonstrating the potential of vibroacoustic design in meeting modern performance standards. Interview: Vanessa Cool – From Structure To Sound: Unlocking The Potential Of Vibroacoustic Design This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.designforam.com

    21 min
  6. AUG 11

    Constrained Creativity In AI-Accelerated Automotive Design - Ruben Verhack - Datameister

    Recorded at CDFAM Computational Design Symposium, Amsterdam , 2025 https://cdfam.com/amsterdam-2025/ https://www.datameister.ai The presentation explores the transformative potential of AI in automotive design, emphasizing the need for tools that embrace constrained creativity within the industry’s unique challenges. Current generic 2D and 3D generative tools often fall short in addressing the intricate constraints of automotive design, such as manufacturability, stakeholder needs, and engineering requirements. By collaborating with a world-renowned design studio, we have mapped out a novel inside-out design process that integrates custom AI tooling directly into designers’ workflows. Rather than replacing designers, this approach empowers them by eliminating non-creative, time-consuming tasks and reducing design iterations between stakeholders. The result is a faster, more effective path to optimal designs that balance creativity, feasibility, and client requirements. During the talk, we will share insights from this collaboration and showcase in-house results that highlight how AI can redefine workflows in automotive design. Interview: Datameister – Constrained creativity in AI-accelerated automotive design https://cdfam.com/datameister-constrained-creativity-in-ai-accelerated-automotive-design/ https://www.datameister.ai This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.designforam.com

    19 min
  7. AUG 4

    Open Source CDFAM - Aaron Porterfield - F=F Presentation at CDFAM Amsterdam, 2025

    Recorded at CDFAM Computational Design Symposium, Amsterdam , 2025 https://cdfam.com/amsterdam-2025/ Open Source, Parametrics, and Lattice Obsession https://fequalsf.blogspot.com/p/about.html https://www.food4rhino.com/en/app/crystallon In this deeply personal and wide-ranging talk, industrial designer Aaron Porterfield reflects on his path from early explorations with BASIC and Blender to founding F=F (Form Equals Function). Aaron shares how a childhood without video games led to a lifelong fascination with design tools, parametric modeling, and open-source collaboration. From early contributions to the Blender Foundation and experiments with lattice hinges and origami tessellations to impactful work on custom 3D-printed medical devices, this talk traces the evolution of his design thinking. Aaron walks through highlights including: Founding the Crystallon plugin for Grasshopper Collaborating on patient-specific helmets and orthotics Volunteering at Burning Man with Arthur Mamou-Mani Building a DIY compression testing rig for lattice experiments Rediscovering and reviving open-source projects with community help He concludes with a discussion on creative commons, the value of attribution, and how open-source culture helps people like him—self-proclaimed “severely ADD” designers—get their projects finished… by others. This presentation captures the spirit of CDFAM: openness, experimentation, and the computational rethinking of design at every scale. 📍 To learn more about the CDFAM Computational Design Symposium, visit cdfam.com #computationaldesign #additivemanufacturing #opensource #grasshopper3d #blender3d #latticestructures #parametricdesign #medicaldevices #burningman #crystallon This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.designforam.com

    15 min

About

CDFAM Computational Design Symposium Presentation Recordings www.designforam.com