I'm a CS and Math sophomore at Berea College, where I work on the Student Software Development Team maintaining two production Flask applications used across the college's Labor Program by 1,500+ students and supervisors. I'm also part of Break Through Tech AI through Cornell Tech, focused on applied machine learning. Outside of Berea, I mentor Ethiopian high school students in algorithms with AddisCoder.

The thing I care most about is what happens to code after you leave. The codebase I work in has 4,000+ commits and a developer team that turns over every few years, so most of my work is about leaving things easier to pick up than I found them. I'm writing a literature review on the same question: why long-lived software systems decay, and what keeps them from it.

Currently: rebuilding AddisChat with retrieval over a real document corpus and an evaluation harness that measures whether the answers are actually grounded.

Exploring: retrieval-augmented generation and evaluation for LLM systems

Open source, Berea College Student Software Development Team

Two Flask applications used across Berea's Labor Program, serving 1,500+ students and supervisors. I shipped a contract management feature end to end, fixed the team's broken Docker dev container setup, and wrote pytest integration tests against a live MySQL instance. Seven years of history, 4,000+ commits, and a developer team that turns over every few years.

Solo project

I wrote my own 30-item questionnaire, distributed it to secondary school students, and cleaned 134 real responses full of inconsistent free-text and missing values. Trained a Random Forest classifier, caught target leakage and rebuilt the analysis, then found that a 5-question subset held 0.663 accuracy under stratified 5-fold CV — 83% shorter survey for a 2-point accuracy cost. Every respondent got their predicted style back with matched study strategies.

Solo project — Best Project, Kotebe University of Education

A chat application built to help students at my school understand how GPT models actually work under the hood. Won Best Project at Kotebe University of Education's Science Shared Campus. I'm rebuilding it now with retrieval over a real document corpus and an evaluation harness to check whether the answers are grounded.

• Software Developer, Student Software Development Team, Berea College — Flask applications for the college Labor Program
• AI Fellow, Break Through Tech AI via Cornell Tech, applied machine learning through eCornell
• Algorithms Mentor, AddisCoder, teaching Ethiopian high school students
• Fellow, Headstarter
• Math TA, developmental mathematics, Berea College
• Bonner Scholar, Berea College — Pugsley Mathematics Award, 2026

Outside of school and code: chess, and FIFA with friends back home.