RAHUL DEBNATH

Built and compared prediction models using Decision Tree, XGBoost, SVM & more. Applied VIF-based multicollinearity reduction, Bayesian networks & DoWhy for causal inference on risk factors.

SVM, KNN, Logistic Regression, Random Forest & CNN for promoter sequence classification. One-hot encoding for ATCG nucleotide sequences. Achieved 95.45% test accuracy with CNN.

A zero-dependency, browser-native tool that turns any audio file into a living 3D sculpture of sound. Drop in a song — watch it become a glowing path through a four-dimensional latent space, computed entirely in your browser in real time. No server, no installs, no API keys. Built with raw Web Audio API + Three.js custom shaders.

Custom image processing to estimate clothing dimensions. Defined Valgrad, deep grad & side grad functions. Regression to convert 2D image measurements into 3D garment parameters. Under Prof. Bimal Kumar Roy & Prof. Debashis Sengupta, ISI.

Analyzed significant shifts in consumer expenditure using Lorenz Curve and Engel Curve techniques. Developed schemas for merging complex unstructured datasets. Under Prof. Bimal Kumar Roy, ISI Kolkata.

Survey-based project assessing public confidence in electoral process. Analyzed voter priorities in Bonhoogly-Baranagar area, comparing 2024 vs 2019 data. Under Prof. Jayant Jha & Prof. Debashis Sengupta, ISI.

Prepared a wrapped presentation of Spotify playlist analysis findings using SQL and animated visualization techniques. Deep dive into music consumption patterns and trends.

Survey to capture public perception of poverty indicators across regions of West Bengal. Focused on contrastive comparison and perturbative performance thresholds. Under Prof. Bimal Kumar Roy & Debashis Sengupta.

Comprehensive comparison of statistical and machine learning models across different problem types. Evaluating performance metrics, bias-variance trade-offs, and selection criteria.