Arpit Kapoor

Environmental Data Scientist

Arpit Kapoor
Get In Touch Download CV

Hi! I am a Postdoctoral Research Associate at the Faculty of Engineering, University of Sydney, Australia. I am a part of computational water-resource modelling research group led by Professor Lucy Marshall (Deputy Vice-Chancellor of Community and Leadership). I completed my PhD in Machine Learning and Hydrology, titled "Process-aware hybrid deep learning for environmental systems and extremes" from the School of Mathematics and Statistics, University of New South Wales (UNSW), in May 2026, where I was supervised by A/Prof Rohitash Chandra, Dr. Sahani Pathiraja, and Professor Lucy Marshall.

During my PhD, I was a part of the HDR cohort of ARC Training Centre in Data Analytics for Resources and Environment (DARE Centre), where I underwent rigorous training in data science and resource modelling. Additionally, I have contributed to the bias correction efforts for climate projections at the Bureau of Meteorology, where I worked in the part-time role of Climate Research Support Scientist for 2 years, modernising legacy workflows.

My research interest lies in the synergy of physics-based hydrological modelling and machine learning to advance environmental process modeling. I have developed hybrid and Bayesian deep learning approaches for rainfall–runoff modeling, flood forecasting, cyclone prediction, and groundwater flow emulation.

Primary Research Interests
  • Bayesian Machine Learning
  • Hydrology & Water Resources Modelling
  • Physics-informed Machine Learning
  • Deep Learning

Professional Experience

University of Sydney

Postdoctoral Research Associate

MAR 2026 - PRESENT

Focusing on developing advanced machine learning methods for environmental process modeling, including operator learning and Bayesian approaches for uncertainty quantification in hydrological systems.

Australian Bureau of Meteorology, Sydney

Climate Science Support Officer

FEB 2023 - APR 2025

Implemented multivariate bias correction method (MRNBC) for the Australian Climate Service. Developed Python wrapper for FORTRAN-based bias correction and implemented parallel compute framework using Dask on NCI Gadi for large-scale climate datasets.

Quince, Hyderabad

Data Scientist-2

MAR 2022 - AUG 2022

Improved customer retention by identifying key drivers of repeat behavior. Built and deployed regressive tree models for predicting and optimizing logistics costs in online retail. Interpreted model predictions using Explainable AI methods (SHAP and LIME).

3Qi Labs, Hyderabad

Data Scientist

NOV 2019 - NOV 2021

Worked extensively with machine learning and data mining tools on Big Data technologies. Developed Deep Learning models for data quality issues detection including referential integrity failure and outlier detection using TensorFlow, Elasticsearch, and Hadoop.

Bomotix, Hyderabad

Machine Learning Developer

JAN 2019 - NOV 2019

End-to-end development and deployment of Deep Learning pipelines in Apache MxNet using Kubernetes and Docker. Specialized in computer vision problems including object detection, tracking, and human pose estimation.

Technical Skills

Python & Machine Learning
Deep Learning (TensorFlow, PyTorch)
Bayesian Machine Learning
Climate & Environmental Modeling
Big Data Technologies (Hadoop, Spark)
Cloud Computing (AWS, Docker, Kubernetes)
Data Visualization (Tableau, Matplotlib)
Scientific Computing (FORTRAN, R)

Featured Projects

Key Research Projects

DeepGR4J

DeepGR4J

Hybrid approach combining conceptual hydrological model GR4J with CNN and LSTM for improved streamflow prediction.

Code Paper
VaRNN

Variational RNN

Variational Bayes algorithm implementation for RNN models applied to cyclone track and intensity prediction.

Code Paper
Parallel Tempering

Parallel Tempering for Neuroevolution

Multi-core implementation of parallel MCMC for population-based Bayesian neuroevolution in deep learning.

Code Paper

Additional Research Projects

BNTL
Bayesian Neural Transfer Learning

Uncertainty quantification in transfer learning

Quantify uncertainty in Transfer Learning with Bayesian Neural Networks trained via MCMC.

Surrogate
Surrogate-assisted Parallel Tempering

Accelerated Bayesian inference

Surrogate models to estimate computationally expensive objective functions in Parallel Tempering.

Bayeslands
Surrogate-assisted Bayeslands

Landscape evolution modeling

Surrogate-assisted tempered MCMC for Bayesian inference in landscape topography evolution models.

RL Humanoid
Humanoid Maze Solver

Deep reinforcement learning

Hierarchical RL approach for teaching maze-solving to humanoid robots using MuJoCo simulation.

Selected Publications

QDeepGR4J: Quantile-based ensemble of deep learning and GR4J hybrid rainfall-runoff models for extreme flow prediction with uncertainty quantification

Journal of Hydrology (2025)

Extension of DeepGR4J using quantile regression-based ensemble learning framework for uncertainty quantification in streamflow prediction and extreme flow event identification.

Publisher PDF
DeepGR4J: A deep learning hybridization approach for conceptual rainfall-runoff modelling

Environmental Modelling & Software, 169, 105831 (2023)

A novel hybrid approach combining process-based hydrological models with deep learning for improved streamflow prediction.

Publisher PDF
Cyclone trajectory and intensity prediction with uncertainty quantification using variational recurrent neural networks

Environmental Modelling & Software (2023): 105654

Implementation of variational Bayesian methods for uncertainty quantification in cyclone prediction using RNNs.

Publisher PDF
Bayesian neuroevolution using distributed swarm optimization and tempered MCMC

Applied Soft Computing (2022): 109528

Novel approach combining swarm optimization with tempered MCMC for training Bayesian neural networks.

Publisher PDF
Surrogate-assisted parallel tempering for Bayesian neural learning

Engineering Applications of Artificial Intelligence, 94, 103700 (2020)

Accelerated Bayesian inference using surrogate models to reduce computational overhead in neural network training.

Publisher PDF
Bayesian neural multi-source transfer learning

Neurocomputing, 378, 54-64 (2020)

Uncertainty quantification in transfer learning using Bayesian neural networks with MCMC sampling.

Publisher PDF

Awards & Achievements

Key Awards & Recognition

Alan Turing Institute

Data Study Group Facilitator

The Alan Turing Institute, UK

May 2024

IROS Challenge

Third Position

IEEE/RSJ IROS Humanoid Application Challenge

Vancouver, Canada - September 2017

Robogames

Multiple Medals

International Robogames Humanoid League

San Francisco, USA - April 2017

Academic & Professional Recognition

PhD Research Training Grant

ARC Training Centre in Data Analytics for Resources and Environments (DARE)

University International Postgraduate Award

Full scholarship for PhD studies at UNSW Sydney

Research Internship Grant

3-month funded internship at University of Sydney

Information Resilience PhD School 2024

Selected participant, Brisbane, Queensland

WAT Consult Best Innovation Award

Atmos Techfest, BITS Pilani Hyderabad Campus - 2017

Engineering the Eye 5 Hackathon

Shortlisted participant, LVPEI Hyderabad (MIT Media Labs) - 2016

Get In Touch

I'm always open to discussing new opportunities, research collaborations, or interesting projects in data science and machine learning.

Email

arpit.kapoor@sydney.edu.au

Location

Sydney, Australia

×