A search engine focused on Wikipedia pages. Creates embeddings for text and images using open source CLIP models, stored and queried via a vector database. All components are containerized with Docker for modular management.
An iOS app for identifying plants, animals, and fungi from photos. Trained a visual transformer on 10,000 species using PyTorch on an Nvidia A100 GPU. Also includes a geolocation model that predicts nearby species from lat/lng coordinates.
An iOS app that performs semantic searches on photos in your photo library using a CLIP neural network.
A JAX-based Python neural network library. Allows creating, training, and saving neural networks with a clean API. Available on PyPI.
Predicts future stock returns from historical returns by outputting a normal distribution over expected future performance.
A working currency and security exchange for items in my apartment. Implements a full order book for buying and selling securities with the custom currency.
A basic web crawler written in Rust, built as a prerequisite to the Wikipedia search engine project.
A C program that encodes any file into a DNA sequence. Motivated by future applications in biological data storage — when DNA writing matures, this could store arbitrary files inside living organisms.
An attempt at writing an operating system from scratch in C and assembly. Implements a basic bootloader and screen driver.