CS senior at Taylor University — AI/ML concentration, Spanish minor.

I work at the intersection of computational linguistics and LLMs: tools that let language communities do technical work without having to write code.

keyboard-studio/keyboard-studio · TypeScript · React · Vite · Vercel

A browser-based authoring studio that turns a linguist's plain-language survey answers into a compiling Keyman keyboard — no code required. Built at SIL International for language communities that need a keyboard for their writing system.

100 commits, 4th of 8 contributors. What I own in it:

• Draft persistence — localStorage + server-backed (Vercel Blob / Postgres) snapshots, so an author can close the tab mid-survey and resume on another device
• My Keyboards — multi-project dashboard for authors juggling several keyboards at once
• GitHub delivery pipeline — authenticated fork → commit → pull request against the upstream keyboards repo, in one click. Built both the user-token (OAuth/PKCE) and org-mediated (GitHub App installation token) paths, plus the serverless token-exchange backend
• Survey wizard reliability — browser-history-synced Back navigation, base-keyboard switching, and resume-state correctness across the multi-step authoring flow
• Engine & validator work — a deterministic simulate() API for testing compiled keyboards, Layer A import-fidelity checks, and DISCUS touch-layout lint rules
• Agent-driven development — most of the above was built by orchestrating multi-agent Claude Code workflows across a TypeScript monorepo

Mini-GPT · PyTorch
Character-level decoder-only transformer (6 layers, 6 heads, 1024-token context) trained from scratch on Shakespeare. Built from Karpathy's nanoGPT walkthrough to learn the mechanics, then extended with my own context-window and training-length ablations.

Churn Prediction · Python · XGBoost · Pandas
End-to-end customer-churn pipeline built for a Kaggle competition: feature engineering, model selection, and evaluation, runnable start to finish from a single entry point.

bible_side · Flutter · Dart
Offline, Material 3 Bible reader built with Freely-Given.org. I'm working on the UI and reading experience through Taylor's Missions Computing Scholar program.

Languages Python · TypeScript · C++ · C · Java · JavaScript · Lisp
ML/AI PyTorch · TensorFlow · XGBoost · RAG · NLP · LLMs · Model Context Protocol · multi-agent systems
Tools React · Vite · Flutter · Pandas · Git · Vercel · Playwright · Vitest

Teaching Assistant for Problem Solving (Python) · Cybersecurity CTF team · Spanish (conversational) · Missions Computing Scholar