Salva Rühling Cachay

Salva Rühling Cachay
salvaruehling at gmail dot com

I am a first-year PhD student at UC San Diego, advised by Prof. Rose Yu. I was a research intern at Palo Alto Research Center (PARC) in summer 22 (and continue being affiliated as a visiting researcher) working with Dr. Kalai Ramea at the intersection of climate modeling and machine learning (ML). Before, I had the pleasure to be researching at Mila - Quebec AI Institute, with Prof. David Rolnick on climate change+ML.

I got started in ML research through a project on weakly supervised learning with Ben Boecking and Prof. Artur Dubrawski at Carnegie Mellon University's Auton Lab. I also led a diverse team of fellow undergraduates in our research project on improving El Niño forecasts with graph neural networks, in terms of both performance and interpretability. We were supported by a Microsoft AI for Earth grant ( see our spotlighted profile).

I am German-Peruvian, and was privileged to be able to grow up in both countries and that volunteering/work enabled me to have a wonderful time living in Colombia, India, and Canada for extended periods of time. I enjoy immersing myself into nature and new cultures, travelling, and exotic fruits. As I now live in sunny San Diego, surfing will probably be on that list soon too!

CV / GitHub / Google Scholar / LinkedIn / Twitter

News

October 22: Organizing a workshop on climate+ML & giving an oral presentation at the Graduate Climate Conference in Pack Forest, WA!
August 22: I was awarded a Jacobs School of Engineering Fellowship :)
August 22: Accepted to the sponsored NASA Summer School on Satellite Observations and Climate Models in Pasadena, CA!
June 22: Started a research internship at Palo Alto Research Company (PARC) on climate+ML!
June 22: Presenting our ClimART work as a contributed talk at the Helmholtz-Zentrum Hereon Data Science Symposium in Geesthacht, Germany!

Research

My goal is to develop and use machine learning methods for positive real-world impact in areas like climate science, climate change, and sustainability. On the machine learning side, I am particularly interested in self-supervised learning, graph structure learning, spatiotemporal forecasting, and conditional generative modeling (including diffusion models, of course :).

ClimART: A Benchmark Dataset for Emulating Atmospheric Radiative Transfer in Weather and Climate Models
Salva Rühling Cachay*, Venkatesh Ramesh*, Jason N.S. Cole, Howard Barker, David Rolnick
NeurIPS Track on Datasets and Benchmarks, 2021
Also: NeurIPS 2021 Tackling Climate Change with ML workshop (Spotlight)
arXiv / OpenReview / code / poster / slides /

End-to-End Weak Supervision
Salva Rühling Cachay, Benedikt Boecking, Artur Dubrawski
Neural Information Processing Systems (NeurIPS), 2021
arXiv / code / poster / slides /

The World as a Graph: Improved El Ni˜no Forecasting with Graph Neural Networks
Salva Rühling Cachay, Arthur F. Bucker, Willa Potosnak, Ernest Pokropek, Emma Erickson, Suyash Bire, Salomey Osei, Björn Lütjens
Under review at IEEE TNNLS, 2021
Also: NeurIPS 2020 Tackling Climate Change with ML workshop
arXiv / code / poster / slides / talk /

Dependency Structure Misspecification in Multi-Source Weak Supervision Models
Salva Rühling Cachay, Benedikt Boecking, Artur Dubrawski
ICLR Weakly Supervised Learning workshop (Oral), 2021
Also: NeurIPS 2020 LatinX in AI workshop (Oral)
arXiv / poster / slides / talk /

Credits to Leonid Keselman and Jon Barron