Special shoutout to @sachalewin.bsky.social for a huge code refactor + super clean GitHub 𝐰𝐢𝐤𝐢.
👉 Explore the repo, try Appa (weights on Hugging Face huggingface.co/datasets/mon...), and let us know what you think!

Big thanks to everyone who made this possible 🙏 @sachalewin.bsky.social, @francois-rozet.bsky.social, @orochman.bsky.social, Victor Mangeleer @mapirlet.bsky.social, Elise Faulx, Marilaure Grégoire and our promoter @glouppe.bsky.social for his tremendous support.

Next stop: 𝐒𝐚𝐧 𝐃𝐢𝐞𝐠𝐨! ✈️
We’ll present Appa at the workshop and continue improving:
• Incorporating real observations
• Reducing temporal error growth
… and we’d love your suggestions or feedback!

We believe Appa carries important messages. 🌍
How to flexibly incorporate observations to tackle diverse assimilation tasks in a 𝐮𝐧𝐢𝐟𝐢𝐞𝐝 𝐟𝐫𝐚𝐦𝐞𝐰𝐨𝐫𝐤? What are the advantages of compressing data for inference at this scale? And what are the limitations?

This release builds on our recently submitted ML4PS @ NeurIPS workshop paper, and already goes beyond the first arXiv version (April 2025, arxiv.org/abs/2504.18720).
Appa now uses a simpler architecture and still reaches strong performance with ~𝟓𝟑𝟎× 𝐜𝐨𝐦𝐩𝐫𝐞𝐬𝐬𝐢𝐨𝐧!

🚀 After more than a year of work — and many great discussions with curious minds & domain experts — we’re excited to announce the public release of 𝐀𝐩𝐩𝐚, our latent diffusion model for global data assimilation!

Check the repo and the complete wiki!
github.com/montefiore-s...

Positions remain open! Both PhD and postdoctoral opportunities are available on scientific foundation models. An additional position is also available on AI for regional climate models (jointly with @xavierfettweis.bsky.social). Do not hesitate to apply!

📢 I am looking for AI post-doc/research/engineer positions in Europe (Paris, London, Zurich, ...) starting 2026. My work revolves around generative modeling and AI for Science, with 4+ publications at top conferences during my PhD. If you are hiring, please reach out! If not, please repost 🔁

Andry, Rozet, Lewin, Rochman, Mangeleer, Pirlet, Faulx, Gr\'egoire, Louppe: Appa: Bending Weather Dynamics with Latent Diffusion Models for Global Data Assimilation https://arxiv.org/abs/2504.18720 https://arxiv.org/pdf/2504.18720 https://arxiv.org/html/2504.18720

Kudos again @sachalewin.bsky.social for the visualizations!

montefiore-sail.github.io/appa/

After several month, our work is out! 🦬

Look at our new paper with the team!

Thanks to all of them!!

@glouppe.bsky.social
@sachalewin.bsky.social
@francois-rozet.bsky.social
@orochman.bsky.social
@mapirlet.bsky.social
Victor Mangeleer
Elise Faulx

Let's start this workshop @ecmwf.int about data assimilation! 🌍

🔥 Azula 0.3.0 is out with many new features, including new super fast samplers (LMS, DPM-Solver), new guidance methods (DPS, MMPS) and new plugins (VDM, EMD, EMD2). Check it out!

Now accepted at @tmlr-pub.bsky.social 🥳

What channels would you recommend for sharing PhD/post-doc position offers? The theme is foundation models for the physical sciences, with weather as a specific target.

📣 Hiring! I am looking for PhD/postdoc candidates to work on foundation models for science at @ULiege, with a special focus on weather and climate systems. 🌏 Three positions are open around deep learning, physics-informed FMs and inverse problems with FMs.

I can't breath anymore but it was so nice!

So much work """just""" for an autoencoder... 🤯😵

Exploring data assimilation with score-based models—and as a bonus, I know what a Markov blanket is!

Score-based Data Assimilation by @francois-rozet.bsky.social and @glouppe.bsky.social

I'm interested too 🤓

Whoops 😅