cv

Data Science Summer Institute Intern @ libROM Team

Lawrence Livermore National Laboratory

Collaborated with interdisciplinary teams to develop data-driven models for large-scale metal 3D printing, working across domains including scientific machine learning, additive manufacturing, and fluid dynamics.

• Performed large-scale simulations on 112 HPC cores with Ansys Mechanical for additive manufacturing data
• Developed a parameterized graph convolutional autoencoder for predicting distortion
• Developed a POD-GP reduced-order model (ROM), and benchmarked it against the graph autoencoder
• Applied Gaussian Process Regression to quantify uncertainty in the ROM predictions
• Predicted distortion within 1e-3 mm in LPBF simulation and achieved speed-up by 1800 times using POD-GP

MITACS Accelerate Intern

Clear Seas

Developed real-time underwater noise prediction models using physics-informed machine learning for marine environmental research.

• Leveraged high-performance computing (HPC) to generate large-scale acoustic datasets using BELLHOP solver
• Developed a range-conditional convolutional neural network (RC-CAN) for predicting transmission loss
• Introduced a continual learning framework for training dynamical systems with varying bathymetry

Graduate Research Assistant @ AI Team

Computational Multi-physics Lab, UBC

Conducting advanced research in machine learning models for spatial-temporal data, focusing on convolutional autoencoders and attention-based neural networks.

• Developed convolutional autoencoder and attention-based RNN for time-series forecasting, improving performance by 5x compared to RNN-LSTM
• Presented results at international conferences and published findings in the Artificial Intelligence issue of Physics of Fluids
• Built a space-time ROM utilizing a 3D convolution kernel for fluid flow prediction, published in Physics of Fluids

Undergraduate Research Assistant

Autonomous Ground Vehicle Research Lab, IIT Kharagpur

Led the development of vision modules and algorithms for autonomous vehicles, with a focus on deep learning and image processing.

• Developed a new vision module using deep learning to upgrade image processing methods, improving computational efficiency and FPS
• Devised a lane detection algorithm using machine learning, improving accuracy by 10%
• Published lane-detection algorithm in IEEE conference on Control, Automation, Robotics, and Vision
• Led team to 2nd place in the Autonomous Navigation Challenge at Oakland University, USA

Doctor of Philosophy (PhD)

University of British Columbia

Mechanical Engineering

• Machine Learning Fundamentals
• Deep Learning for Computer Vision
• Numerical Methods for Deep Learning
• Deep Learning with Graph
• Introduction to Probability
• Computational Fluid Dynamics

Bachelor of Technology (BTech)

Indian Institute of Technology Kharagpur

Aerospace Engineering

• Algorithms and Data Structures
• Deep Learning
• Machine Learning
• ML for Medicine

Top Three Presenters

Lawrence Livermore National Laboratory

Recognized as one of the top three presenters at the Data Science Summer Internship summer slam program.

Department GSI Award

University of British Columbia

Awarded department GSI award for hard work and outstanding research contributions in 2022 & 2023.

International Tuition Award and Graduate Research Fellowship

University of British Columbia

Awarded for outstanding research at UBC.

Department Rank 2

IIT Kharagpur

Secured department rank 2 among all undergraduate students in B.Tech Aerospace Engineering.

Shastri Student Internship Project Grant

Shastri Indo-Canadian Institute

Awarded grant among thousands of applicants from India and Canada.

1st Position in Smart India Hackathon-2020

IIT Kharagpur

Led the team to 1st position among 60 teams in IIT-Kharagpur round of Smart India Hackathon-2020.

Harnessing Loss Decomposition for Long-Horizon Wave Predictions via Deep Neural Networks

NeuralIPS ML4Physics Workshop

Exploring loss decomposition techniques to improve long-horizon predictions using deep neural networks.

Data-Driven, Parameterized Reduced-order Models for Predicting Distortion in Metal 3D Printing

NeuralIPS ML4Physics Workshop

Collaborative work on parameterized reduced-order models for metal 3D printing, highlighting efficiency in predicting distortion.

Continual Learning of Range-Dependent Transmission Loss for Underwater Acoustics Using Conditional Convolutional Neural Net

arXiv

Research on employing continual learning in neural networks for transmission loss predictions in underwater acoustics.

A Multi-Objective Optimization Framework for Reducing the Impact of Ship Noise on Marine Mammals

Ocean Engineering

Developed a multi-objective optimization framework to mitigate the impact of ship noise on marine mammals.

Combined Space-Time Reduced-Order Model with 3D Deep Convolution for Extrapolating Fluid Dynamics

Physics of Fluids

Developed a space-time reduced-order model for fluid dynamics simulation using a 3D convolutional neural network, selected as Editor's Pick.

Hypergraph Message Passing Network for Fluid Dynamics Simulations

Developed a hypergraph message passing network to enhance fluid dynamics simulations, achieving improved time prediction compared to existing models.

Deep Learning for Extrapolating Fluid Dynamics

Conceptualized and implemented a temporal convolutional neural network (CNN) for extrapolating beyond the training regime, improving time-forecasting accuracy by 15%.

Graduate Teaching Assistant

University of British Columbia

Delivered tutorials and lab sessions for Data Science 100, Introduction to Probability, and Statistical Inferences courses at UBC, contributing to the education of over 300 students.

Team Head

Autonomous Ground Vehicle Lab, IIT Kharagpur

Managed a student design team of 40 UG students, developing vision modules for autonomous vehicles and organizing weekly meetings to ensure project milestones were met.