§ Projects
Things built, tested, and written up.
A working list of agentic AI and software-engineering projects. When a project turns into a paper, the publication is attached to the project.
- 01
Mozilla × University of Calgary · 2026 · Paper in preparation
Firefox QA Test-Plan Agent
An agentic GPT-4 workflow that drafts candidate QA cases, compares them against human-written Firefox test plans, and surfaces missed coverage for human triage.
PublicationUsing LLMs to Bridge the Gaps in QA Test Plans at Firefox
- Agentic AI
- GPT-4
- QA
- Firefox
- 02
Empirical study · 2025 · Submitted paper
Class-Code Prompting Bench
A benchmark for measuring how prompt strategy shapes the quality of class-level code generated by large language models, tested against compile-and-test correctness.
PublicationPrompting Matters: Assessing the Effect of Prompting Techniques on LLM-Generated Class Code
- LLM
- Code generation
- Prompting
- 03
Outreachy × Mozilla Firefox · 2024 · Shipped in open source
Bugbug Model Contributions
Production ML contributions to Mozilla's Bugbug pipeline, including accessibility-bug tagging, spam-comment classification, and training workflow observability.
- Mozilla
- ML
- Bugzilla
- MLOps