AYUSH RAJ

A mind on Earth. An eye on what’s next.

AYUSH RAJ

ML RESEARCH · WEATHER & CLIMATE

I build machine-learning systems for weather and climate.

Fifth-year BS–MS student at IISER Pune, currently researching subseasonal-to-seasonal precipitation forecasting with deep learning and ERA5 data.

Open to PhD opportunities and machine-learning roles.

Explore my research Read my CV
Ayush Raj BS–MS · IISER Pune Pune, India About my work

FIG. 01 / A WORLD OF POSSIBILITIESIllustrative model · not live data

Current MS thesis · subseasonal forecasting
Education BS–MS · IISER Pune
Methods Deep learning · HPC

01 / Research & engineering

Scientific questions.
Practical ML.

Machine learning, Earth-system science, and scientific computing.

My work connects scientific questions with the code and data needed to investigate them. My current focus is weather and climate, building on earlier research in quantitative medical imaging.

RESEARCH DIRECTION

Questions I want to explore.

  • Learning atmospheric patterns for weather and subseasonal-to-seasonal forecasting.
  • Understanding how extreme heat forms, persists, and spreads.
  • Developing reproducible computational methods for scientific data.
Research experience & collaborators

ENGINEERING EXPERIENCE

Tools behind the research.

Model training
PyTorch · FSDP · Multi-GPU
Scientific data
ERA5 · Xarray · DeepGraph · HPC
Quantitative imaging
SciPy · NiBabel · SimpleITK
Explore my code on GitHub

02 / Selected projects

Complex systems.
Real-world questions.

Weather forecasting, climate analytics, and quantitative imaging.

01 / Weather AI Featured research

FuXi weather forecasting

Reproducing and scaling a global AI weather model using 40 years of atmospheric data.

Question
How can a global AI weather model be reproduced and trained across decades of atmospheric data?
My contribution
Worked on FuXi reproduction and distributed training using PyTorch, FSDP, and ERA5.
Research context
Weather and climate research at IISER Pune with Dr. Bedartha Goswami.
  • PyTorch
  • FSDP
  • ERA5
  • Multi-GPU

View project repository

02 / Climate analytics

Spatio-temporal heatwave atlas

Finding where extreme heat forms, persists, and spreads across long-term climate records.

Question
Where does extreme heat form, persist, and spread across long-term climate records?
My contribution
Worked on spatio-temporal heatwave analysis using ERA5 data, Xarray, and DeepGraph.
Research context
Weather and climate research at IISER Pune with Dr. Bedartha Goswami.
  • Xarray
  • DeepGraph
  • HPC

View project repository

03 / Medical imaging

Quantitative MRI pipelines

Building reproducible pipelines that make quantitative MRI measurements more stable.

Question
How can quantitative MRI measurements be made more stable and reproducible?
My contribution
Built IVIM–DKI fitting and denoising workflows for DICOM and NIfTI imaging data.
Research context
Research internship with IIT Delhi and AIIMS, working with Dr. Amit Mehndiratta.
  • SciPy
  • NiBabel
  • SimpleITK

Read research experience

Forecast anatomy Atmospheric state → ML forecast → scientific evaluation
Reference levels · relative pressure · not live data
250 hPaUpper-tropospheric flow
500 hPaMid-troposphere
850 hPaLower troposphere
1000 hPaNear-surface reference
  1. 01 / Represent ERA5 reanalysis

    Forty years selected from global ERA5 reanalysis.

  2. 02 / Learn Deep-learning models

    Scalable training across atmospheric variables.

  3. 03 / Evaluate Forecast skill

    Careful comparison across lead times and regions.

03 / Research path

Follow the
curiosity.

Current work and selected research collaborations.

  1. Current MS Thesis Researcher Subseasonal rainfall forecasting SCDLDS · Ashoka University

    Developing deep-learning methods for subseasonal-to-seasonal precipitation forecasting with ERA5.

    • ERA5
    • Deep learning
    • Forecast evaluation

    With Dr. Sandeep Juneja

  2. Research project Weather & Climate Research Forecasting and heatwave analysis IISER Pune

    Worked on FuXi forecasting, distributed training, and spatio-temporal heatwave analysis across ERA5 data.

    • FuXi
    • ERA5
    • Distributed training

    With Dr. Bedartha Goswami

  3. Research internship Quantitative MRI Research Stable, reproducible imaging pipelines IIT Delhi and AIIMS

    Built reproducible IVIM–DKI fitting and denoising workflows for DICOM and NIfTI medical imaging data.

    • IVIM–DKI
    • DICOM / NIfTI
    • Denoising

    With Dr. Amit Mehndiratta · Remote

PhD opportunities · ML roles · Research collaborations

Pune, India · IST (UTC+5:30)

LET’S TALK

I’m interested in PhD opportunities and machine-learning roles where scientific questions and practical model development meet.

Discuss a PhD opportunity Discuss an ML role
raj.ayush@students.iiserpune.ac.in