Artificial Intelligence undergraduate focused on building verifiable AI systems, reproducible LLM evaluation, local-first software, and practical AI applications.
- AI systems and agent workflows with explicit validation boundaries
- Reproducible evaluation of LLM agent reliability and feedback adaptation
- Local-first applications that keep users in control of their data
- Practical and industrial AI prototypes with limitations stated clearly
A local-first Windows sticky-note application built with Tauri, Rust, React, TypeScript, and SQLite. It includes a mobile web companion, self-hosted synchronization, automated tests, CI, and a downloadable Windows release.
The current MVP is designed for personal use and trusted local networks; its documentation explicitly covers the absence of end-to-end encryption and other production security boundaries.
An industrial AI prototype and case study for exploring image-derived heuristic features, material matching, and visualization. It is a demonstration system, not a production-validated waste-steel grading model.
Reproducible evaluation of LLM agent reliability and feedback adaptation, with engineering verification kept separate from research-hypothesis evidence.
Python · TypeScript · React · Rust · SQLite · Tauri · LLM evaluation · testing · reproducible experiments
Evidence before claims. Reproducibility where results matter. AI-assisted engineering with human review, tests, and explicit product and research boundaries.