About Me
Hello! I’m Mushrafi Munim Sushmit. Welcome to my page. I am a master’s student in the Department of Physics at the University of Dhaka. My research explores the intersection of physics and quantum computing, unraveling the mysteries of the universe.
I’m currently working on my Master’s thesis, exploring the innovative application of neural networks to quantum computing. This work blends classical neural network models with the vast capabilities of quantum computing. Before diving into my Master’s research, I spent time as a research assistant, investigating everything from disease dynamics influenced by vaccination decisions to the use of quantum machine learning for predicting solar irradiance.
If you’re interested in the convergence of physics and technology or looking for a collaborator who brings a blend of passion, knowledge, and practical experience to the table, let’s connect!
Interests
- Artificial Intelligence
- Quantum Computing
- Physics Informed Neural Networks
- Nonlinear Dynamics and Chaos
Education
MSc in Physics, Present
University of Dhaka
BSc in Physics, 2023
University of Dhaka
Skills
Programming Language
Python
C++
Julia
Bash
Technical Skills
Quantum Computing
Quantum Machine Learning
Mathematical Modeling
Data Science
Machine Learning
Computer Vision
Natural Language Processing
Frameworks
Nix
Git
PyTorch
Qiskit
Comsol
Pandas
OpenCV
Linux
Research Experience
Adapting Physics-Informed Neural Networks for Quantum Computing
Master’s Thesis
- Conducting research under the mentorship of Dr. Golam Dastegir Al Quaderi in the field of quantum computing, specifically focusing on the adaptation of physics-informed neural networks (PINNs) to quantum computational frameworks.
- Successfully developed multiple models that integrate classical physics equations with PINN, including the coupling of differential equations like the Lotka-Volterra model, high-dimensional coupled Higgs equation, and nonlinear Schrödinger equations.
- Currently transitioning towards quantum machine learning-based implementations, aiming to bridge the gap between classical physics-informed modeling and quantum computing capabilities.
- Utilizing a combination of Python, PyTorch, PennyLane, and Qiskit to construct and validate quantum-enhanced versions of physics-informed neural networks.
- This ambitious project, serving as my master’s thesis, aims to pioneer the integration of PINNs within quantum computing, potentially revolutionizing how we approach and solve physics-based problems through computational means.
Vaccination Decisions in a Dual Strain Disease Dynamics
Research Assistant
- Worked under the guidance of Dr. Muntasir Alam from Department of Applied Mathematics, University of Dhaka to analyze the impact of vaccination considering behavioral, socio-economic factors on the dynamics of a second disease strain.
- Integrated randomness, game theory principles, and network analysis to create a comprehensive model for disease spread in the given scenario.
- Utilized Python, Mesa and Julia to implement multi agent based models and conducted rigorous testing.
- Developed three distinct models to enhance the understanding of disease spread dynamics: a scale-free model where agents utilize a random scale-free graph to make vaccination decisions, a complete graph model reflecting uniform decision-making among all agents, and a stochastic model for validating our findings.
- Published the findings in a Q1 journal with an impact factor of 7.8.
Quantum Machine Learning for Solar Irradiance Forecasting
Research Assistant
- Worked under the supervision of Dr. Mohammed Mahbubul Islam from Institute of Energy Engineering, Dhaka University of Engineering & Technology to develop and validate hybrid classical-quantum machine learning models for solar irradiance prediction.
- Conceptualized and initiated the project, proposing the integration of quantum layers within feedforward neural networks.
- Utilized Python, PennyLane, Qiskit, and PyTorch to implement the models and conducted rigorous testing.
- Engineered a novel fully connected parameterized quantum circuit tailored for solar irradiance forecasting.
- Led the technical aspects of the project, encompassing model design, implementation, model validation and, performance evaluation.
- Published the findings in a Q1 journal with an impact factor of 10.4.
Accomplishments
Quantum Computing theory, Comsol Simulation of Quantum Mechanical Systems, IBM Qiskit Learning
See certificate
Neural Networks & Deep Learning
Foundational concepts of neural networks, Deep learning architectures
See certificateBayesian Methods for Machine Learning
Bayesian Networks, Markov Chain Monte Carlo, Bayesian Inference
See certificateData Driven Astronomy
Python Programming, Machine Learning, Applied Machine Learning, SQL
See certificateProjects
*
My Personalized NixOs Flake
This Repository contains some useful bash scripts i have written for day to day use
This repository features a simulation framework for modeling disease spread utilizing game theory to understand decision-making processes in infectious disease transmission.
This repository contains implementations of Physics-Informed Neural Networks (PINNs) to solve various physical systems described by partial differential equations (PDEs) or ordinary differential equations (ODEs). PINNs leverage the power of deep learning to approximate solutions to complex physical and quantum mechanical problems by incorporating physical laws as part of their loss function.
This repository is dedicated to advancing Automatic Speech Recognition (ASR) for the Bengali language, leveraging state-of-the-art machine learning models such as wav2vec 2.0, T5, ARPA, BERT, and BART. This project is part of an experiment to understand and improve ASR performance in processing and recognizing Bengali speech, aiming to create more accurate and efficient ASR systems for Bengali, the seventh most spoken language in the world.
This is an animation of a Julia set using the Numba and ffmpeg in a Jupyter-notebook environment.
Publications
(2024).
Dynamic vaccination strategies in dual-strain epidemics: A multi-agent-based game-theoretic approach on scale-free hybrid networks.
Chaos, Solitons & Fractals.
(2023).
Forecasting solar irradiance with hybrid classical–quantum models: A comprehensive evaluation of deep learning and quantum-enhanced techniques.
Energy Conversion and Management.