Bas Veeling

🚀We're excited to emerge from stealth and announce our $30M Seed financing round. A huge thank you to our incredible investor group including @HoxtonVentures, @BasisSet, @lightspeedvp, @northzoneVC, @localglobevc, Touring Capital, Giant Ventures, @FjLabs, @ZeroPrimeVC & Tiferes Ventures who all share and back our vision. We’ve set out on a mission to transform how materials are designed and developed using AI and, in the process, tackle some of society’s most critical challenges. We recognise that the need for such developments could not be greater and have assembled a world-class team motivated by this very fact. To further accelerate our impact, we're immensely grateful to have the support of @geoffreyhinton who has joined our advisory board. We are also excited to announce our partnership with @ylecun and the @Meta FAIR team to develop advanced materials for carbon capture - a critical piece of the climate puzzle. We're just getting started, but you can learn more about our work in the full press on our site. @wellingmax // @ac_edwards_1 fortune.com/europe/2024/06… #AI #machinelearning #climatechange #sustainability

Announcing our new model for materials! FlowMM... - Generates stable & novel materials efficiently - Predicts crystal structure accurately - Generalizes Riemannian Flow Matching to point clouds w/ periodic boundaries arxiv.org/abs/2406.04713 @RickyTQChen @anuroopsriram @bwood_m

We have a new distillation method that actually *improves* upon its teacher. Moment Matching distillation (arxiv.org/abs/2406.04103) creates fast stochastic samplers by matching data expectations between teacher and student. Work with @emiel_hoogeboom @JonathanHeek @tejmensin. 1/4

We introduce Geometry-Informed Neural Networks to train shape generative models without any data (!!), combining learning under constraints, neural fields as a suitable representation, and generating diverse solutions to under-determined problems: 🖥️: arturs-berzins.github.io/GINN/

Excited to introduce Aurora: a foundation model of the atmosphere.  In <1min, Aurora produces 5-day global air pollution predictions and 10-day high-resolution weather forecasts that outperform SOTA classical simulation tools and the best specialized deep learning models… 1/n

I'm super excited to finally share what we've been working on for a long time: Aurora, a foundation model of the atmosphere.

@wellingmax Congrats Max!

Thank you Yisong and the Award Committee for choosing the VAE for the Test of Time award. I like to congratulate Durk who was my first (brilliant) student when moving back to the Netherlands and who is the main architect of the VAE. It was absolutely fantastic working with him.

@dpkingma Congrats Durk!

Thanks to the ICLR Award Committee! And thank you for the kind words, Max! You were the perfect Ph.D. advisor and collaborator, kind and inspiring. I really couldn't have wished for better.

At #ICLR2024 presenting our latent protein simulation poster with the amazing @mfederici_! Tuesday at 10:45, Halle B #64 Also happy to discuss proteins sampling, latent sim., and graph contrastive rep. learning, hit me up. iclr.cc/virtual/2024/p…

[1/7] 🎉Our paper Time-lagged Information Bottleneck (T-IB) is in #ICLR2024! Kudos to @BasVeeling @ryotat Patrick Forré & MSR AI4Science! 📄arxiv.org/abs/2309.07200 T-IB maps complex dynamics to simple latent spaces for super fast, highly accurate simulations!

@HassanConor @vdbergrianne @wellingmax Hi @HassanConor! I did not, but happy to chat if you have any questions!

@BasVeeling @vdbergrianne @wellingmax Hey @BasVeeling , this looks super interesting! Did you ever release an implementation of this work? Thanks

Tired of gaussian variational posteriors? Quantize your latent variables for more flexible posteriors! It improves predictive uncertainty, learns non-linearities, doesn't require normalization and scales well 🎉. At arxiv.org/abs/1810.05500. With @vdbergrianne @wellingmax

[1/N] Generative AI has revolutionized how we create text and images. How about designing novel materials? We at @MSFTResearch #AI4Science are thrilled to announce MatterGen: our generative model that enables broad property-guided materials design. 👇 arxiv.org/abs/2312.03687

Sharing an early preprint of my Microsoft AI4Science summer internship project. We developed SE(3) flow matching for protein backbone generation. Compared to SE(3) diffusion, we find our method achieves higher designability, faster sampling, with a way simpler implementation. 1/8

Interested in accelerating scientific discovery with AI? Join our interdisciplinary global team! Two exciting opportunities for engineers: Amsterdam or Berlin: jobs.careers.microsoft.com/global/en/job/… Beijing: jobs.careers.microsoft.com/global/en/job/…

Do you want to obtain accurate, long-horizon predictions with neural PDE solvers? Introducing PDE-Refiner, a training process that improves rollout accuracies, offers uncertainty estimates and better data efficiency. 📜: arxiv.org/abs/2308.05732 🖥️: phlippe.github.io/PDERefiner/ 🧵1/9

Personal news: I have recently joined @MSFTResearch AI4Science in Berlin. Very excited to work with @FrankNoeBerlin and the fantastic team!

Super exciting research engineering position on the interface between AI and material discovery @MSFTResearch AI4Science!

This is the paper I should have written. #NeurIPS2022 #workboat2022

🤯🤯🤯 #NeurIPS2022

New internship openings at AI4Science @MSFTResearch in Cambridge (UK), Amsterdam and Berlin! careers.microsoft.com/us/en/job/1497…