Personalized Visual Analytics for Multi-Criteria Decision Making
March 2025 – Present
University of Zürich
Identified shortcomings in current multi-criteria decision systems that lead to suboptimal, non-transparent rankings
Developed an interactive VA framework to capture complex user preferences via explicit inputs and implicit feedback
Engineered a prototype with advanced attribute scoring, uncertainty visualization, and LLM-powered natural-language explanations in a human-in-the-loop system
Established a scalable, transparent prototype improving decision accuracy and user trust for real-world applications
Evaluating the Stability of SHAP Under Class Imbalance
January 2024 – October 2024
University of Leeds
Developed and executed an experimental framework to assess the stability of SHAP values across machine learning models (SVM, KNN, Decision Trees, XGBoost) using imbalanced datasets
Developed SMOTE-Tomek hybrid sampling to create datasets with varying class imbalance ratios (1%, 5%, 10%, till 50%)
Introduced a novel metric, Coefficient of Variation (CV), for quantifying SHAP value stability
Generated visual insights using SHAP plots and box plots to communicate findings on feature importance stability
SHAPXAISMOTE-TomekPythonMachine Learning
Fourier Analysis and Roth's Theorem
July 2021 – July 2022
IISER Bhopal
Studied Fourier analytic techniques and applications to Roth's Theorem in additive combinatorics
Explored recent advancements in the finite field setting and breakthroughs in upper bounds for the Cap-set problem using the polynomial method
Developed understanding of mathematical proofs in additive combinatorics and the elegance of solving problems