Welcome to my GitHub portfolio! I am Rashika Gupta, a data scientist and Nuclear Physicist. Currently, I am busy wrapping up my thesis to achieve the milestone of "Dr." My work spans advanced simulations, deep learning, and interdisciplinary applications in healthcare, clean energy, and more. Below you'll find a curated selection of my projects, my PhD thesis, and contributions that reflect my journey and expertise.

Jump to: PhD Thesis | Projects | Certifications

Fav quote:

"A theory that you can't explain to a bartender is probably no damn good." - Rutherford

Title: Simulations of Electron Detection Systematic Errors for a UCNA+ Experiment

• Abstract: The UCNA+ experiment at Los Alamos National Laboratory is an upgrade to the original UCNA experiment, aiming to measure the neutron beta asymmetry A0 with a precision of 0.2% or better, nuclear structure-independent value for the CKM matrix element Vud. In this study, we present results from GEANT4 simulations of the UCNA+ experimental setup, focusing on necessary backscattering corrections.
• Keywords: Electron Backscattering, Geant4 Simulations, Detector Physics, Data Analysis, UCNA+, MCMC

Read the full thesis here.
Conference Presentation here.
Code here.

• Overview: Leveraged on pretrained models to learn from skin cancer images to predict if it was melanoma.

• Status:

• Models Used: SeResNext model.

• Results: Results showed significant improvement over baseline models in capturing field variations.

• Technologies:

• Overview: Tackled a multiclass classification problem on a heavily imbalanced Intel dataset, using techniques like SMOTE and PCA for dimensionality reduction. The models, such as XGBoost and RandomForest, were optimized through hyperparameter tuning and cross-validation.
• Status:
• Results: Achieved an F1-score of 0.978 using XGBoost after feature reduction and oversampling techniques.

• Overview: Developed and applied Physics-Informed Neural Networks (PINNs) to model complex fluid dynamics scenarios. This project integrates deep learning with physical laws to enhance simulation accuracy.

• Status:

• Models Used: Custom Physics-Informed Neural Network (PINN) architecture.

• Results:

  • Successfully solved the Navier-Stokes equations.
  • Achieved high accuracy in predicting fluid flow fields compared to traditional CFD methods.

• Technologies:

• Overview: This project involves the simulation of electron backscattering and energy deposition using Geant4. The insights gained contribute to improved detector design and understanding of particle interactions.

• Status:

• Models Used: Geant4 simulation models, ROOT for data analysis.

• Results:

  • Successfully simulated energy deposition profiles at varying depths.
  • Analysis of material impact on backscattering efficiency led to optimized detector designs.

• Technologies:

• Mentorship: Guided junior researchers in their projects, focusing on simulation techniques and deep learning applications.
• Collaboration: Worked with interdisciplinary teams at Los Alamos National Lab, contributing to precision experiment reducing back-scatering by 50%.

• MLOps & Cloud Computing: Completed certifications in MLOps, AWS, and related cloud technologies.
• Deep Learning Specialization: Advanced courses in deep learning, focusing on neural networks, optimization techniques, and unsupervised learning.

View full list of certifications here.

Feel free to reach out if you're interested in collaborating, discussing research, or exploring new opportunities!

• Email: [email protected]
• LinkedIn: Rashika Gupta

Thank you for visiting my GitHub portfolio! 😊