Karl D.D. Willis retweeted
New #ICCV2025 paper: ✨ Aligning Constraint Generation with Design Intent in Parametric CAD ⚙️
We apply post-training techniques to the task of generating engineering sketch constraints found in parametric CAD, using a constraint solver for verifiable rewards.
At #ICML this week @SamXu03799145 will present our latest work that generates solid CAD models in a controllable manner. Check out the tweet below for details!
In our latest #ICML2023 paper "Hierarchical Neural Coding for Controllable CAD Model Generation" hnc-cad.github.io/, we learn the design intent from different CAD hierarchies in a self-supervised manner, and demonstrate its application on multiple tasks e.g. autocomplete.
The ICML Poster session is Thu 27 Jul 2023 at 10:30 a.m. HST — noon HST: icml.cc/virtual/2023/poster/…
Karl D.D. Willis retweeted
Latest interview in the CDFAM speaker series with @KarlDD on his AI research at Autodesk. lnkd.in/grBZKVPP
Karl will be speaking more about his work at the Autodesk Research AI Lab at CDFAM NYC June 14-15.
Register to attend, today (May 15th) is th…lnkd.in/g8yYfJ5Q
Interested in an internship at Autodesk Research? We have three ML roles for Summer 2023 in SF and Toronto.
Data Acquisition and ML Research
linkedin.com/posts/karldd_in…
Large Language Models
linkedin.com/posts/karldd_in…
Machine Learning Research
linkedin.com/posts/karldd_au…
Karl D.D. Willis retweeted
An automated way to assemble thousands of objects: A new algorithm for automatic assembly of products is accurate, efficient, and generalizable to a wide range of complex real-world assemblies. mitsha.re/i2tu50LZyLn
Our latest collaboration with @MIT_CSAIL and @TAMU led by @YunshengTian will be presented at SIGGRAPH Asia today!
We want to make it easier to assemble objects AND to disassemble them for end-of-life recycling and part reuse.
Assemble Them All: Physics-Based Planning for Generalizable Assembly by Disassembly
abs: arxiv.org/abs/2211.03977
project page: assembly.csail.mit.edu/
For a nice overview of our work, check out this article from MIT: csail.mit.edu/news/automated…
If you are at #NeurIPS2022 this week, check out some of the work the Autodesk AI Lab team has been doing on robustness, evaluating graph generative models, and communicating natural programs!
Autodesk Research AI Lab has 3 papers accepted at #NeurIPS2022!
✅ MaskTune: Mitigating Spurious Correlations.. bit.ly/3UhpVlT
✅ Evaluating Graph Generative Models.. bit.ly/3ubQf6z
✅ Communicating Natural Programs to Humans & Machines bit.ly/3VDwcto
Karl D.D. Willis retweeted
AI-powered design tools are essential to Autodesk’s future offerings, and the work being published by the AI Lab at top-tier AI conferences gives us a glimpse into the future of design.
bit.ly/3zhP9ZM
Karl D.D. Willis retweeted
want to get into program synthesis but don't know how to started? I wrote a minimalist intro to modern program synthesis that can help you -- from problem formulation to generating code by fine-tuning llm on huggingface.
evanthebouncy.github.io/prog…
Large language model failure mode of the month: Mode collapse to all emoji output all the time. Not helpful 🤷♂️😫🤬 @evanthebouncy
Collage is a whole lot of fun! Chatting to @_joelsimon and @ev_ancasey was even more so. Super cool/smart team making progress on the grand challenge of how we interact with AI to create.
We’re hard at work on a new way to make generative art, and are excited to share our first prototype!
Create any scene you can imagine with a simple collage and a prompt.
Try it out! collage.artbreeder.com/
📢 We are hiring in the Autodesk AI Lab. Doubling down on what @adam_gaier said, for all those folks that are mission driven to create tools for designing better buildings, cars, devices etc and want to use AI/ML to make it happen.
Come work with me at Autodesk!
i'll pick you up at the intersection of AI and design
apply here ⬇️
autodesk.wd1.myworkdayjobs.c…
Fun collaboration between the Autodesk AI Lab and @jiajunwu_cs lab at Stanford! Learning to assemble (and also disassemble) is both technically challenging and impactful for manufacturing efficiency and ensuring products don't have a negative impact at end-of-life.
For folks who are still at #ICML2022 I'll be giving a talk today at the Workshop on Machine Learning for Computational Design. Exciting line up!
mlcompdesign.github.io
icml.cc/virtual/2022/worksho…















