David Levin
@diwlevin.bsky.social
520 followers 91 following 270 posts
Sometimes professor sometimes nvidia research scientist all-the-time loud mouth Opinions are my own and do not represent either of my employers
Posts Media Videos Starter Packs
diwlevin.bsky.social
The NVIDIA HiFi team (research.nvidia.com/labs/prl/) is starting our search for summer interns. Are you demonstrably awesome at physics simulation, robotics, scene reconstruction, high-performance numerical methods? DM me your CV.
NVIDIA High Fidelity Simulation Research
research.nvidia.com
diwlevin.bsky.social
If you are an instructor and you'd like access to the solution repo, DM me, I'm happy to share.
diwlevin.bsky.social
The old assignments in C++ were tricky for students because UofT doesn't teach it in a lot of courses. I'm hoping moving to Python means people can concentrate more on learning concepts rather than syntax.

Setup using pip is much easier as well, and they can optionally run on the GPU using CUDA.
diwlevin.bsky.social
This year I'm reworking my physics-based animation course with new assignments in Python (using Torch and NVIDIA Warp). So far it's been a great set of tools to use. For the interested, here's the student code for the first assignment: github.com/dilevin/pba-...
GitHub - dilevin/pba-assignment-fem: Implementing the volumetric finite element method using PyTorch and NVIDIA Warp
Implementing the volumetric finite element method using PyTorch and NVIDIA Warp - dilevin/pba-assignment-fem
github.com
diwlevin.bsky.social
When you use A.I. for fishing it’s called vibe codding
diwlevin.bsky.social
My new number one use for LLMs is, I kid you not, spell and grammar checking

Somehow ChatGPT is much better at correcting my often typo ridden drafts than whatever MS Word is doing.

Turns out if you ask it to make only the minimal required changes, it will abide.
diwlevin.bsky.social
New research direction !!! AI really is a trillion dollar market
diwlevin.bsky.social
You got to learn about XYZ, that’s more informative then these form letter exchanges used to be :)
diwlevin.bsky.social
Sitting in the SIGGRAPH Asia EC town hall.
Meant to be about LGBTQ+ concerns

Currently we are being given tourist info about Malaysia

Starting to think that the whole EC is either incompetent or just doesn't care about the opinions of their constituents.
diwlevin.bsky.social
Maybe you just need more people ?
Reposted by David Levin
cemyuksel.com
I am boycotting #SIGGRAPHAsia 2026 because it will be in Malaysia, where there are serious safety concerns for the LGBTQ+ members of our community, thereby excluding them. I will not attend, submit my work to, or perform any reviews for SIGGRAPH Asia 2026.

Yet, I'm excited about #SIGGRAPHAsia 2025!
diwlevin.bsky.social
Below is my response to being invited to serve on the technical papers committee at SIGGRAPH Asia 2026 (in Malaysia)

TLDR It's a no.
Dave writes email
diwlevin.bsky.social
Not really a comment on quality of talks and more a comment on my own particular mental make up.
diwlevin.bsky.social
I think my ideal technical talk length on a single topic is 15 minutes. Enough time to get a good sketch of the method but not so much that my attention wanders. After that I really just need to sit with the paper to have any hope of understanding.
diwlevin.bsky.social
It’s the name of my kraftwerk cover band
diwlevin.bsky.social
SIGGRAPH over

Tired now

Or Maybe hungry

Off to go solve this mystery.
diwlevin.bsky.social
This was really fun. Thanks to all the speakers. Would do again, so hopefully SIGGRAPH keeps the workshop program around.
diwlevin.bsky.social
Me,Peter Chen and Eitan Grinspun are hosting a SIGGRAPH workshop on reduced-order models. Great keynotes from Maurizio Chiaramonte (Meta), Jernej Barbic (USC) and Mengyu Chu (Peking University) plus a bunch of awesome lightning talks.

Come check it out: roms.news
ROM Workshop at SIGGRAPH 2025
Explore the future of subspace and neural simulation techniques at SIGGRAPH 2025. Hear from top researchers and join the conversation on model reduction.
roms.news
diwlevin.bsky.social
This morning 9am East Building Room 11
diwlevin.bsky.social
Me,Peter Chen and Eitan Grinspun are hosting a SIGGRAPH workshop on reduced-order models. Great keynotes from Maurizio Chiaramonte (Meta), Jernej Barbic (USC) and Mengyu Chu (Peking University) plus a bunch of awesome lightning talks.

Come check it out: roms.news
ROM Workshop at SIGGRAPH 2025
Explore the future of subspace and neural simulation techniques at SIGGRAPH 2025. Hear from top researchers and join the conversation on model reduction.
roms.news
diwlevin.bsky.social
Happening tomorrow morning !
diwlevin.bsky.social
Me,Peter Chen and Eitan Grinspun are hosting a SIGGRAPH workshop on reduced-order models. Great keynotes from Maurizio Chiaramonte (Meta), Jernej Barbic (USC) and Mengyu Chu (Peking University) plus a bunch of awesome lightning talks.

Come check it out: roms.news
ROM Workshop at SIGGRAPH 2025
Explore the future of subspace and neural simulation techniques at SIGGRAPH 2025. Hear from top researchers and join the conversation on model reduction.
roms.news
diwlevin.bsky.social
This is probably why no one likes the idea to deanonymize the reviews
diwlevin.bsky.social
SIGGRAPH Asia 2025 preemptively boycotted me … 0 for 4 on submissions this year.

Well played universe !
diwlevin.bsky.social
It’s like the thing with the decimal point

Europe just does find/replace ./, :)
diwlevin.bsky.social
Oh you “cogitated” over this human rights issue ? Well that solves it.

Sheesh …