My Projects
Project Title
Short description of the project.
Demo Video
Additional Information
Here you can provide more details about the project, its features, and any other relevant information.
NBA Team Performance Modeling with Hidden Markov Models
- Developed a probabilistic agent that models the sequential evolution of NBA team performance using HMMs to predict game outcomes.
- Implemented Baum-Welch training and Viterbi decoding in Python to infer hidden team performance states and predict win probabilities.
Social Media Comment Classifier
- Worked in a group following standard ML pipelines and built Naïve Bayes and SVD Logistic Regression models using Scikit-learn to classify social media comments as toxic.
- Extracted textual features using TF-IDF to represent text data for downstream classification models.
Recipe Card Site(Let me Cook)
- Collaborated in a 10-person team to design and deploy a recipe website with user uploads, filtering, and saving features using Agile methods.
- Engineered dynamic DOM manipulation routines and state-managed filtering algorithms for seamless recipe searching.
Link to Github
Stock Price Event Analysis
- Analyzed the impact of negative exogenous events on sectoral stock prices using hypothesis testing and regression analysis.
- Researched and designed data models to compare sector-level and market-level trends.
Chess AI
- Developed a UCI-compliant C++ chess engine (compiled to WASM) along with a demo site performing roughly around 2500 Elo.
- Implemented alpha-beta pruning, transposition tables, bitboard representations, magic bitboards(for O(1) move generation) and much more features.
Link to Github
Connect 4 AI
- Built a benchmarking site with several different algorithms and implementations to note amount of improvement and significance from each feature and optimization.
- Engineered bitboard state representations, heuristic evaluations, different search functions to optimize connect 4 solver, benchmarking node evaluations across algorithmic variations.
Link to Github
Full-Stack Web Analytics & Reporting Engine
- Developed a full-stack web analytics platform featuring multi-tier Role-Based Access Control (RBAC) across Super Admin, Analyst, and Viewer permissions.
- Engineered relational MySQL database schemas and Express routes to log user activity, generating dynamic data visualisations using Chart.js and automated PDF export reports.
- Engineered an automated CI/CD pipeline using GitHub Actions to continuously deploy updates via SSH to a multi-vhost Apache server hosted on an Ubuntu instance.
(Not live at the moment)
Generative Music & Audio Synthesis Model
- Built an unconditioned note language model using a pitch-token Gated Recurrent Unit (GRU) in PyTorch to generate symbolic MIDI music sequences.
- Processed a dataset of ~1,599 classical piano excerpts; quantized note timing into 120-tick steps and converted compositions into pitch-token sequences across a 128-pitch vocabulary (including BOS/EOS/PAD tokens).
- Partitioned dataset into a 90/10 train/validation split by file and trained the GRU model as a lightweight, expressive alternative to Markov chains and Transformers.