Ayush Raj
BS–MS, IISER Pune · MS thesis at Ashoka University
I am a final-year BS–MS student at IISER Pune, currently completing my master’s thesis at the Safexpress Centre for Data, Learning and Decision Sciences (SCDLDS), Ashoka University.
My research focuses on machine learning for weather and climate prediction, particularly regional subseasonal-to-seasonal rainfall forecasting over India. I also work on global AI weather models and the spatio-temporal analysis of extreme heat.
I am advised by Prof. Sandeep Juneja and collaborate with Dr. Manmeet Singh and Prof. Parthasarathi Mukhopadhyay, with research guidance from Dr. Bedartha Goswami.
I also contribute to NVIDIA Earth2Studio, where my FuXi-S2S model integration has been merged, and work on S2S Research, an experimental platform for evaluating forecast skill over India.
News
- Two first-author papers accepted for poster presentation at the NeurIPS 2026 Tackling Climate Change with Machine Learning workshop (CCAI). Publications
- Speaker at the 3rd Annual Workshop on AI/ML Methods in Weather and Climate Modelling, hosted by SCDLDS at Ashoka University.
- Started my MS thesis research on subseasonal-to-seasonal rainfall forecasting at Ashoka University.
- Began a transformer-based weather forecasting project at IISER Pune with Dr. Bedartha Goswami.
- Joined the Centre for Biomedical Engineering at IIT Delhi as a summer research intern in quantitative MRI.
Publications
IndiaS2S-Bench: A Comprehensive Benchmark of Subseasonal to Seasonal Forecasting over India
Tackling Climate Change with Machine Learning (CCAI), NeurIPS 2026 Workshop
Accepted for poster presentation
Calibrating AI Subseasonal to Seasonal (S2S) Forecasts with Lightweight Regional Adapters
Tackling Climate Change with Machine Learning (CCAI), NeurIPS 2026 Workshop
Accepted for poster presentation
Research
Improving Regional Subseasonal-to-Seasonal Forecasting
Master’s thesis · SCDLDS, Ashoka University
Advisor: Prof. Sandeep Juneja
Developing deep-learning methods for regional subseasonal-to-seasonal precipitation forecasting over India using ERA5 data.
FuXi weather forecasting
IISER Pune · with Dr. Bedartha Goswami
Reproducing and scaling a global AI weather model across 40 years of atmospheric data. Code
Spatio-temporal heatwave analysis
IISER Pune · with Dr. Bedartha Goswami
Studying where extreme heat forms, persists, and spreads using ERA5, Xarray, and DeepGraph. Code
Quantitative magnetic resonance imaging (MRI)
Summer research internship · Centre for Biomedical Engineering, IIT Delhi and AIIMS
Supervisor: Prof. Amit Mehndiratta
Developed reproducible fitting and denoising workflows for quantitative MRI during a summer research internship.
Open-source contributions
Contributor · NVIDIA Earth2Studio
FuXi-S2S integration
Contributed the FuXi-S2S prognostic model for global subseasonal forecasting, including ensemble support, tests, documentation, inference examples, and GPU validation.
Research software
Research platform
S2S Research
An experimental platform that evaluates subseasonal forecast skill over India and highlights the best-performing forecast for each region.
WebsiteContact
For research correspondence: raj.ayush@students.iiserpune.ac.in