evan
@evan__casey
Followers
872
Following
5K
Media
165
Statuses
1K
computational tools for visual art, design, animation researcher at autodesk ai lab, creator of @cadmiumai
Porto, Portugal
Joined July 2011
New #ICCV2025 paper:Β β¨ Aligning Constraint Generation with Design Intent in Parametric CAD βοΈ We apply post-training techniques to the task of generating engineering sketch constraints found in parametric CAD, using a constraint solver for verifiable rewards.
1
4
5
Stop by our poster at ICCV today from 3:15-5:15pm π Special thanks to my collaborators at @ADSKResearch: Tianyu Zhang @shuishidaOx @wp_mccarthy John Thompson @AKhasahmadi Joseph Lambourne and Pradeep Jayaraman @karldd
0
0
0
Check out the paper, blog post, website for more results! π Paper: https://t.co/k3yHmiaDXI Blog post: https://t.co/0fc5RC7u7U Project website:Β https://t.co/gA3ELNIzOa
research.autodesk.com
Explore the next step in AutoConstrain, focusing on ensuring...
1
2
1
Our method outperforms pure supervised approaches by a large margin, with online RL approaches (GRPO/RLOO) showing the largest performance gains. We jointly optimize for fully-constrained sketches while also minimizing geometric distortion and over-constrained sketches.
1
0
0
Only a fraction of user-generated sketches are fully-constrained, so pre-training + SFT can only get so far. To address this, we apply post-training techniques using verifiable rewards from a constraint solver.
1
0
0
Constraints are what make parametric CAD models editable, but theyβre tricky to get right even for experienced CAD designers. A key requirement is that the constraints preserve the π
πππππππβπ ππππππ with respect to salient geometric relationships.
1
0
0
ghost-maxxing is the new bitter lesson pilled
Finally had a chance to listen through this pod with Sutton, which was interesting and amusing. As background, Sutton's "The Bitter Lesson" has become a bit of biblical text in frontier LLM circles. Researchers routinely talk about and ask whether this or that approach or idea
0
0
0
a new tool has entered the room
Excited to announce that Cadmium is out of beta! π€ We have a brand new colorization model available β vastly improved accuracy on complex characters and motion.
2
4
10
We're launching new EbSynth today π EbSynth is VFX software that lets you change your video by editing one frame. You can turn any video into your canvas, right in your browser. Go try it out π
55
131
836
Honestly the most fun and gratifying project I have ever been part of - we're hiring for someone with dev tool chops to make things far out on the pareto frontier of fun + powerful https://t.co/mgfuaPpy30
careers.on.websim.ai
1
6
49
New research project: Lluminate - an evolutionary algorithm that helps LLMs break free from generating predictable, similar outputs. Combining evolutionary principles with creative thinking strategies can illuminate the space of possibilities. https://t.co/lNdFtvjlcm
33
184
1K
when you log on to twitter and your feed is still all big baby vance pics for the 4th day in a row
0
0
4
Some of my early work in AI colorization is now on display in Vienna at Belvedere and outside the Medical University of Vienna. tldr, Gustav Klimt's three Faculty paintings were lost during the second world war, and I collaborated with Google and Belvedere to colorize the
4
2
19
π Announcing the 1RG Incubator: A 12-week program for community-minded technologists. No equity, no feesβjust mentorship, community, and space to build. Building a better Toronto through tech starts locally π± Apply by Feb 15 β¨ https://t.co/5eVcaGGQcJ
#toronto #civictech
incubator.1rg.space
The 1RG incubator is a 12-week mentorship program for individuals who are developing projects at the intersection of technology and community impact.
2
6
8
Introducing Mochi 1 preview. A new SOTA in open-source video generation. Apache 2.0. magnet:?xt=urn:btih:441da1af7a16bcaa4f556964f8028d7113d21cbb&dn=weights&tr=udp://tracker.opentrackr.org:1337/announce
220
565
4K
playing with the new paint mode in @Artbreeder Composer! really neat way to bring drawing into the workflow yet also fully benefit from IPAdapter. you also draw in Collage but it's not as good at retaining likeness in your import images
0
0
3
back to making SVGs with good 'ol backprop for now voronoi grid studies π
2
0
5