Salva Rühling Cachay

PhD student @ucsd_cse. Into AI for Science, especially climate&weather.

[Salva Rühling Cachay @salvaRC7]

Sharing a blog post about our #Neurips2023 paper "DYffusion: A Dynamics-informed Diffusion Model for Spatiotemporal Forecasting"
https://t.co/sYm0ZWL0HZ

It's faster than standard diffusion models, has low memory needs, and generates stable & probabilistic rollout forecasts. 1/7

[Jiahe (Chloe) Huang @jhhuang_chloe]

[🚨ICML 2026 Paper Alert] Excited to introduce SimulCost! 🚀 The first cost-aware benchmark and toolkit for evaluating LLM agents on complex physics simulations.
Can LLMs efficiently tune complex physics simulators under realistic computational budgets? Spoiler: They struggle! 📉
🔗 Project: https://t.co/7zU0FcSn2J
📄 Paper: https://t.co/VNMyYdZgdM
💻 Code: https://t.co/ovecoAhMJ9
🤗 Models: https://t.co/No5cJDmPRb
Huge thanks to the amazing team: @YadiCao, Sicheng Lai, Yang Zhang, Zach Lawrence, Rohan Bhakta, Izzy F. Thomas, Mingyun Cao, Chung-Hao Tsai, Zihao Zhou, Yidong Zhao, Hao Liu, Alessandro Marinoni, Alexey Arefiev, @yuqirose
A quick thread 🧵👇

[Salva Rühling Cachay @salvaRC7]

Together with Duncan Watson-Parris and @yuqirose!

[Salva Rühling Cachay @salvaRC7]

🌎⚡ A frontier (1.5°, 15-day) ensemble weather forecast in ~3 seconds, from a probabilistic model you can train in ~1 day on a single H200 GPU?*
Meet U-Cast, our new #ICML2026 paper. 🧵

[Salva Rühling Cachay @salvaRC7]

*Fine-tuning takes ~1 day (1.2 H200-GPU-days) on a pre-trained deterministic backbone; add ~7 H200-days to train that backbone from scratch (~8 total).
📄 https://t.co/XAHJiG88Yb
💻 https://t.co/pTzKANlPq9

[Salva Rühling Cachay @salvaRC7]

Really appreciate the computational support from @modal !!

[Charles 🎉 Frye @charles_irl]

some people use @modal to predict the next token, others to predict the next rainstorm

https://t.co/IMwIGXHWOQ

[Charles 🎉 Frye @charles_irl]

some people use @modal to predict the next token, others to predict the next rainstorm

https://t.co/IMwIGXHWOQ

[Salva Rühling Cachay @salvaRC7]

NeurIPS poster session tomorrow morning!⚡️

Come say hi if you are interested in diffusion/AI4Science/PDEs/weather/etc.

⏰ Friday, 11am-2pm PST at Poster #3614

Happy to chat outside these hours as well—I'm here for the full week (and the next ;)

[Salva Rühling Cachay @salvaRC7]

A huge thank you to my brilliant collaborators @NVIDIAAI -Miika Aittala, @karsten_kreis, Noah Brenowitz, @ArashVahdat & @MardaniMorteza-and @yuqirose @ucsd_cse

👇 See you next week in San Diego! Paper: https://t.co/T3uOlqnzN2

[Salva Rühling Cachay @salvaRC7]

🌍 Modeling chaos isn't just about predicting the next step—it's about modeling how uncertainty grows over time.🌪️

I’m thrilled to share Elucidated Rolling Diffusion Models (ERDM), accepted to #NeurIPS2025!

We unify rolling diffusion with EDM for forecasting complex systems🧵👇

[Salva Rühling Cachay @salvaRC7]

𝗧𝗵𝗲 𝗥𝗲𝘀𝘂𝗹𝘁𝘀: 3) 𝗣𝗵𝘆𝘀𝗶𝗰𝗮𝗹 𝗥𝗲𝗮𝗹𝗶𝘀𝗺: ERDM matches the power spectra of operational physics-based models (IFS ENS), solving the "blurriness" problem common in AI weather models.

[Dongxia Wu @allen_dongxiawu]

My lab (https://t.co/W2r7eFGkhw) at MBZUAI is recruiting PhD students, Postdocs, and Visiting Scholars starting Fall 2026. Interested candidates can email their CV, transcript, and research plan to [email protected] with the subject: '[Name] [Position] - MBZUAI Application'.