A local, privacy-first Retrieval-Augmented Generation system that lets you chat with your PDF documents using semantic search and LLMs.
A neural network that learns to approximate mathematical functions and images by training a small multilayer perceptron (MLP) to map inputs to outputs with a live visualization that updates every epoch so you can watch the network learn in real time.
A multilayer perceptron implemented from scratch in NumPy, with custom loss functions, optimizers, and backpropagation to gain better fundemental understanding of core DL math.
Transformer encoder model for sentiment analysis built in PyTorch from scratch, for robust text classification.
Convolutional neural network for text classification, inspired by the "CNN for sentiment analysis research paper", exploring kernel sizes (n-grams) and pooling strategies for NLP.
Collection of CNN architectures in PyTorch for image classification experiments and benchmarking.
ASCII animation of a rotating 3D donut rendered in the terminal using linear transformations and projections.
Set of numerical methods to approximate pi, comparing accuracy and convergence across algorithms.
Simulation of the Monty Hall problem to visualize probability and strategy outcomes.