Technical Research & Studies

Submitted to ISMIR 2026

Explainable TCN for AI-Music Detection (Submitted to ISMIR 2026)

Built a lightweight Temporal Convolutional Network (TCN) 482x smaller than baselines, achieving 90.48% accuracy in detecting AI-generated music directly from raw waveforms. Engineered an explainability suite using saliency maps and attention weights to pinpoint generation artifacts.

OnGoing Research

Cross-Border Seagull Trajectory Analysis

Analyzing raw GPS telemetry from Movebank using GeoPandas. Quantifying geospatial mobility through the ReFGeM feature set to identify behavioral differences in cross-border migration patterns.

Retrieval-Augmented Generation for Complex Claims (Python)

Built a Retrieval-Augmented Generation system that processes and retrieves information from a mixture of text and visual data using vector embeddings.

Tim Hortons Simulation (Python)

Discrete-Event Simulation (DES) tool utilizing a Future Event List (FEL) to simulate real-world scenarios. Benchmarks and analyses best staffing and resource allocation strategies.

Input
Output

Ray Tracer (Python)

A ground-up graphics engine simulating photorealistic light transport, shadows, and reflection models.

Input
Output

Linear Program Solver (Python)

Algebraic solver for linear optimization using the Simplex method to find optimal solutions.

Learning Projects

Built by following step-by-step guides to master foundational concepts.

Before Training
After Training

RL: Frozen Lake (Q-Learning)

Training autonomous agents to solve Markov Decision Processes in stochastic environments.

Arduino Simulation

Arduino LED Systems (c++)

Embedded systems design for interactive LED arrays and hardware-level programming.