Faez Ahmed
Massachusetts Institute of Technology, MA·Engineering·Early-career
AI models can now generate images and text, and they could also help engineers come up with and judge designs for cars, aircraft, energy systems and other products. Engineering design is harder for AI than images or text, though.
- Fields
Engineering
Computer Science
- Topics
engineering design
generative AI
deep generative models
- Grants
- 2 known grants, through Feb 2030
What they work on
- Develop diffusion models, generative AI that builds outputs by removing noise step by step, that learn from the full history of design optimization runs rather than only final optimized designs, to generate more precise engineering designs.
- Create algorithms that train deep generative models on both valid designs and invalid designs that fail requirements, so models can learn effectively from smaller engineering datasets.
- Combine parametric, graph and image representations of designs using multimodal contrastive learning, which teaches a model to match different views of the same design, so it can tell similar designs apart without labeled data.
- Compare expert designers, crowd-sourced raters, pure machine learning and expert-assisted machine learning for rating the novelty and quality of CAD drawings, text, sketches and prototypes.
- Measure how design format affects the reliability of crowd ratings, and build validated crowd–AI workflows that evaluate design ideas at scale.
More about the research
Designs must meet strict requirements, good training data is scarce, and the same design can be described in very different formats, such as numbers, CAD drawings, sketches or text. The lab's projects work on two sides of this problem. One builds AI that generates designs. The other develops ways for people and AI together to evaluate how new and how good designs are.
Why it matters
Engineering-focused generative AI could speed up product design across automotive, aerospace, sustainable energy and infrastructure, which could lower costs, allow more customized products and shorten time to market. Reliable ways to score design novelty and quality at scale could help design researchers and companies sort through large numbers of ideas, and the open data could support others working on design evaluation.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Alejandro Ribeiro
University of Pennsylvania, PA·Engineering
Develop optimization and statistical foundations for aligning generative models by exploiting that the problem is strongly concave in distribution space even though it is highly nonconvex in the parameters of deep networks and transformers.
Qiunan Zhang
Alabama State University, AL·Computer Science
Build machine learning models that identify key confounding factors, such as education, location or language, hidden in textual and non-textual user-generated data like online reviews.
Yusu Wang
University of California-San Diego, CA·Computer Science
Explain why trained flow-based and diffusion generative models generalize better than the theoretically optimal solution predicts, using geometry, ODE analysis, manifold learning and deep learning theory, to enable controlled generation of novel samples.
The school
Massachusetts Institute of Technology: 25 labs on PI Pages, $1.1B a year on research, 692 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.