TSP Genetic Algorithm

Used genetic algorithms to optimize the traveling salesman problem over 200 cities.

Temporal Reasoning in POMDPs

Modeled temporal reasoning in belief-state-based partially observable environments using a custom Little Prince simulation.

Sentiment Analysis: TF-IDF + ML

Implemented sentiment classification using TF-IDF vectorization and classical ML models like Naive Bayes and Logistic Regression.

Sentiment Analysis: Word2Vec + Neural Nets

Trained sentiment models using Word2Vec embeddings and implemented neural architectures like feedforward and convolutional networks.

HMM-Based POS Tagging

Built a part-of-speech tagger using Hidden Markov Models with both greedy and Viterbi decoding for inference.

BiLSTM + CNN NER

Built a named entity recognition model using BiLSTM and GloVe embeddings, with optional CNN-based character features.

K-Nearest Neighbors

Implemented KNN classification from scratch with support for Euclidean, Cosine, and Minkowski distances.

Linear & Polynomial Regression

Built regression models to fit linear and nonlinear data, including L2 regularization and training visualization.

Logistic Regression & Perceptron

Trained binary classifiers using gradient descent, stochastic updates, and numerical stability techniques.

Feedforward Neural Networks

Developed a multilayer perceptron from scratch with tanh/RELU activation, dropout, and SGD w/ momentum.

Decision Trees & Boosting

Implemented decision trees and built ensemble models using random forests and AdaBoost for classification.

K-Means Clustering

Clustered data using K-Means and K-Means++ initialization, applied to classification and image compression.

Word Embeddings via PCA

Generated co-occurrence matrices and reduced them using PCA to produce low-dimensional word embeddings.

Hidden Markov Models

Trained HMMs for part-of-speech tagging, implemented forward-backward, posterior, and Viterbi decoding.

Tiny Transformer in PyTorch

Built a decoder-only Transformer from scratch to perform character-level language modeling.

Q-Learning with Experience Replay

Developed a reinforcement learning agent that uses Q-learning and experience replay to learn optimal policies.

Moodtunes

NLP-based system that detects user emotions from text and recommends matching music using lyric and mood similarity.

AI Agents for Little-Go

Built Minimax and Alpha-Beta pruning agents for the Little-Go board game to counter diverse opponent strategies.