News
In the long run, you'd find that perhaps science is far more open-minded than the other religions -- Geoff Hinton
June 2024
Invited talk at the International Center for Theoretical Sciences, Bangalore India in the TAPGFD program. See the recording
March 2024
Invited talk at M2Lines.
January 2024
Invited talk at Amazon Science
December 2023
Contributed talk at the American Geophysical Union
November 2023
Invited talk at the physics-informed machine learning workshop at Raytheon Technologies Research Center
Contributed talk at the Division of Fluid Dynamics, APS
August 2023
Invited talk at the department of mathematics, University of California, Santa Cruz
May 2023
Invited talk at the department of mechanical engineering, University of California, Santa Barbara
March 2023
Invited talk at the Unversity of Texas, Austin in the department of Aerospace Engineering and Engineering Mechanics
February 2023
Invited talk at the Raytheon Technologies Research Center, AI Discipline
Jan 2023
Invited talk in the applied math department at the University of California, Santa Cruz
Invited talk in the atmospheric science department at the University of Wyoming, Laramie
Invited talk at North Carolina State University's Department of Marine, Earth and Atmospheric Sciences
Invited talk at the AI for Earth Systems Insitute on stable deep learning-based digital twins for the Earth. See the recording
New paper on interpretability in deep learning was published in PNAS. See paper. Also a news piece on this work.
New paper on deep learning and data assimilation published in the Journal of Computational Physics. See paper
December 2022
Two talks at AGU 2022 on digital twins and data assimilation
November 2022
Invited talk at NVIDIA Research on long-term stable digital twins for the climate
September 2022
Invited talk at the 2nd Stanford Model Hierarchies Conference
Invited talk at the VESRI Data-Wave Project
April 2022
I delivered a talk at the LBNL, Monterey Data Workshop. You can find the agenda (and possibly the recordings) here.
I delivered a talk at the SIAM Uncertainty Quantification Conference at Atlanta Georgia. The topic is similar to the first part of my talk at the LLNL DDPS seminar. You can find it here.
March 2022
I was invited to speak at the Climate Physics Group at the University of Lausanne (UNIL)-Institute of Earth Surface Dynamics, Switzerland to talk about my analog forecasting work .
We have released a pre-print on our state-of-the-art data-driven weather prediction model, FourCast Net here. This is a giant collaboration between NVIDIA, Rice, Purdue, Caltech, and Berkeley Lab.
Feb 2022
I was invited to speak at the DDPS seminar at the Lawrence Livermore National Laboratory on our recent works at the intersection of data-driven models with physically consistent architectures and integration of data assimilation. Please take a look at the talk here.
Our paper on integrating data assimilation with physically-consistent architectures for data-driven weather prediction is accepted at Geoscientific Model Development. See the paper here.
Our paper on data-driven stable a-posterior LES models for turbulent flow and generalization to high Re with transfer learning has been published in the Journal of Computational Physics. See the paper here.
December 2021
Our work on deep learning to enhance data assimilation has been featured by SIAM news. See here for details.
I was awarded the outstanding student presentation award (OSPA) at AGU 2021 for my work in deep learning-enhanced data assimilation for high-dimensional non-linear systems.
October 2021
We have released a pre-print on our work on discovering model error for chaotic dynamical systems with Bayesian sparse regression. See here.
July 2021
I and Jaideep Pathak at Berkeley lab are organizing a mini-symposium at SIAM Annual Meeting 2021, on Applied Machine Learning for Fluid Physics, MS 99. Please come and join us for exciting talks!
May 2021
New paper on transfer learning, Transfer Learning of Deep Neural Networks for Predicting Thermoacoustic Instabilities in Combustion Systems, published in Energy and AI. See paper.
I and my colleague, Ebrahim, are giving two talks on the theme "deep learning and data assimilation" at SIAM Dynamical Systems 2021, MS 124, Efficient Data Assimilation with Deep Learning-Based Ensemble Generation: Applications to Geophysical Turbulence and MS 169, Data-Driven Forecasting of Extreme Events with Equivariance-Preserving Deep Transformers and Data Assimilation
April 2021
Check out our review paper on Physics Informed Machine Learning for Weather and Climate Modeling published in the Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences
New pre-print at the intersection of data assimilation and deep learning.
March 2021
New Pre-print on developing stable a-posteriori LES models with deep learning for 2D turbulent flow.
Feb 2021
Invited talk in the Advanced Modeling and Simulation seminar series in the department of Mechanical Engineering at University of Texas El Paso, titled " Demystifying the Alchemy: On Theory of Deep Learning and Some Applications in Data Assimilation"
Jan 2021
Invited talk at Berkeley Lab, NERSC's Data Seminar, " Deep Learning Approaches to Modeling Multi-Scale Chaos and Geophysical Turbulence". See the talk here .
Talk at AMS on, " Integrating Data Assimilation with an Equivariance-Preserving Deep Spatial Transformer: Towards Physically Consistent Data-Driven Weather Forecasting"
Dec 2020:
New paper on data-driven subgrid-scale modeling accepted in Physics of Fluids, "Data-driven subgrid-scale modeling of forced Burgers turbulence using deep learning with generalization to higher Reynolds numbers via transfer learning". paper
Nov 2020:
Please check out my talk at APS DFD 2020, in the focus session Deep Learning in Experimental and Computational Fluid Mechanics, titled, "Equivariance-preserving Deep Spatial Transformers for Auto-regressive Data-driven Forecasting of Geophysical Turbulence".
Integrating data assimilation with structurally equivariant spatial transformers: Physically consistent data-driven models for weather forecasting accepted at Neural Information Processing Systems , 2021: AI for Earth Science Workshop
Data-driven super-parameterization using deep learning: Experimentation with multi-scale Lorenz 96 systems and transfer-learning is published in Journal of Advances in Modeling Earth Systems (link)
Oct 2020
Check out my invited talk at ECMWF-ESA Workshop on Machine Learning for Earth System Observation and Prediction (link). Here I talk about my extreme events with analog forecasting and some more recent work at Berkeley lab.
Sep 2020
Aug 2020
Highlights of our work at Berkeley Lab: news !
Paper accepted at the International Conference on Climate Informatics, UK, Oxford. Recommended as one of the top 15% of accepted submissions. Check pre-print . A big shout out to Mustafa and Karthik at Berkeley Lab for their contributions to the manuscript !
Please feel free to check out my invited talk (recording starts from 55:25 min) at the NOAA's 2nd workshop on leveraging AI in environmental sciences. Here I discuss some of the key results from my DDSP work pre-print . There's an interesting panel discussion that follows.
July 2020
I have released a pre-print (link) on the effectiveness of equivariance-preserving deep convolutional encoders for forecasting complex geophysical turbulence. This work was primarily done at Berkeley Lab with Karthik and Mustafa.
May 2020
I start my summer internship at Lawrence Berkeley National Lab with Karthik Kashinath and Mustafa Mustafa. I'd be working as deep learning research intern in the Data and Analytics Services group. My work would be at the intersection of turbulence, deep learning and climate dynamics
deepVein Inc website goes live. We are in business! Check out website.
I received the Emmett T. and Geraldyne Smith Roberts Award from the Mechanical Engineering Dept. at Rice University for demonstrating outstanding quality of research and academic achievements in the Year of 2019-2020.
January 2020
I had a chance to visit my alma mater, Indian Institute of Technology, Patna, India where I delivered a talk in the departmental seminar of Mechanical Engineering Department: On chaos, turbulence and their predictability in the atmosphere: A deep learning approach
December 2019
I delivered a talk in AGU 2019 in San Francisco. It was on one of my favorite works on Capsules
Analog forecasting of extreme-causing weather patterns using deep learning. See the paper: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2019MS001958
Multiple news outlets had covered this work and its potential implication in modern day weather prediction
https://insidehpc.com/2020/02/deep-learning-for-predicting-severe-weather/
https://eos.org/articles/combining-ai-and-analog-forecasting-to-predict-extreme-weather
https://www.xsede.org/-/can-deep-learning-yield-more-accurate-extreme-weather-forecasts-
https://www.theregister.co.uk/2020/02/05/weather_ai_prediction/
https://www.sciencedaily.com/releases/2020/02/200204112518.htm
November 2019
I delivered two talks at APS DFD 2019, in Seattle. While one of them was delivered on behalf of my PI, the other one was mine
Data-driven prediction of multi-scale chaotic Lorenz 96 system: Reservoir Computing, ANN, and RNN-LSTM. See the paper: https://www.nonlin-processes-geophys-discuss.net/npg-2019-61/
Data-driven super-parameterization with deep learning: Experiments with Lorenz 96 system and transfer-learning. See the paper: https://arxiv.org/abs/2002.11167