Hello, my name is Cardell Taylor.

I'm a software engineer and ML researcher.

ABOUT ME

I'm a twenty-one year old programmer, artist, and bread baker and computer science student. I've been writing code since 8th grade. My favorite programming languages are Python 🐍, C++ C++ Logo and Rust 🦀.

Although much of my work is in AI/ML, I'm interested in most things within the realm of computer science. I have some experience working with both front and backend development, Raspbery PIs, and OpenGL graphics programming

I'm currently working in a research lab under Dr. Zachary Debruine, where we develop models to analyze single-cell RNA sequencing data.

GitHub LinkedIn Resume

Projects

Second Mind (WIP)

A local LLM assistant for retrieving information from Obsidian/Markdown notes. It uses retrieval-augmented generation to enable context-aware querying. It supports locally hosting your own LLM model via Ollama for maximum privacy.

Sim-Vest </>

A financial investment simulator written in Rust. Allows you to create portfolios, make trades, and see your performance over time. It uses the Financial Modeling Prep (FMP) API to fetch stock prices in real time.

Capstone - SysML v2 Export Tool

Working in the Agile project framework, assisted Array of Engineers by updating their web app tool for modeling logical requirements. We added the ability to export data in the SysML v2 format, a universal language for modeling system requirements.

Single-cell Integration Research </>

As years go by, the cost of collecting single-cell data gets lower and lower. However, we still lack solid methods for pooling this data together in aggregate. The goal of my research is to develop a new deep learning method to pool together information across several datasets, with an emphasis on the ability to conserve characteristics across species.

MNIST Michi-GAN </>

Exploration of the MichiGAN framework for generating disentangled representations of the MNIST dataset, as detailed in the paper "MichiGAN: sampling from disentangled representations of single-cell data using generative adversarial networks". Performance was evaluated using Fréchet Inception Distance (FID) and Inception Score.

Lean-AE </>

A from-scratch C++ implementation of a variational autoencoder for compressing and reconstructing image data. Implements ADAM to reduce convergence time and uses OpenMP multithreading to boost performance. Later ported to Python to test adversarial feedback techniques.