Piyush Garg
banner
tropmetpie.bsky.social
Piyush Garg
@tropmetpie.bsky.social
Atmospheric scientist navigating the AI tide. Senior AI weather scientist at RWE AI Lab. PhD from UIUC. Chicago resident.
Get ahead of the curve! The official posting is coming soon, but we're eager to connect with talented candidates now. DM to discuss the role or share your profile.
#MachineLearning #ClimateScience #AIResearch #WeatherAI #DataScience #GenerativeAI #ResearchJobs
February 22, 2025 at 3:43 AM
-Track record of applying ML to weather and climate science
-Drive to push the boundaries of environmental AI applications
Bonus expertise:
-Experience with multi-modal weather datasets
-Hands-on work with generative AI and transformer architectures
February 22, 2025 at 3:43 AM
If interested in reading about it in detail, please have a look at our arxiv article here:
arxiv.org/abs/2408.10958
Kilometer-Scale Convection Allowing Model Emulation using Generative Diffusion Modeling
Storm-scale convection-allowing models (CAMs) are an important tool for predicting the evolution of thunderstorms and mesoscale convective systems that result in damaging extreme weather. By explicitl...
arxiv.org
December 6, 2024 at 4:37 PM
Haha this is surely fun! 😂
December 4, 2024 at 3:13 AM
Yes it is indeed a large array with quite a good number of channels and thus we went with zarr. I think zarr is of course more optimized for large tensors but the benefit also shows up in smaller tensors as they are easily scalable data format, similar to HDF.
December 1, 2024 at 3:43 PM
We have been experimenting with NVIDIA DALI for deep learning training and inference work with Pytorch and zarr-enabled data loaders. It really has optimized our training and inference pipeline.

github.com/NVIDIA/DALI
GitHub - NVIDIA/DALI: A GPU-accelerated library containing highly optimized building blocks and an execution engine for data processing to accelerate deep learning training and inference applications.
A GPU-accelerated library containing highly optimized building blocks and an execution engine for data processing to accelerate deep learning training and inference applications. - NVIDIA/DALI
github.com
November 30, 2024 at 9:18 PM
Would be happy to be a part of it! 🙂
November 24, 2024 at 10:35 PM
I wish to specially thank @mikepritchard.bsky.social for giving me the opportunity to work with and learn from his team at NVIDIA. I surely have learnt so much during one year of my postdoc tenure here at NVIDIA. 🙏🏽
November 21, 2024 at 5:38 PM
Will be happy to be on the list! 🙂
November 21, 2024 at 1:37 AM
There definitely are a bunch of ML based air quality models out there right now. I believe Microsoft’s Aurora also has that capability if you want to give it a try! 🙂
November 17, 2024 at 1:54 AM
Thank you Katharine! That sounds wonderful! 🙏🏽
November 15, 2024 at 1:23 PM
Hey Nathanael! I guess it was about time 😂
November 13, 2024 at 2:07 AM