AI engineer · Software QA / verification and validation · Australia
I work on AI evaluation, API verification and practical software tools. My background is in automotive test engineering and cloud.
| Project | What it does | What to inspect |
|---|---|---|
| agent-eval-harness | Compares coding models and agent commands on the same tasks, using withheld pytest suites. | Task definitions, grading code and reproducible result records. |
| api-vv-toolkit | Generates API tests from OpenAPI contracts and written requirements, with a traceability matrix. | Test generation, requirement mapping and report outputs. |
| Rambler Bangla | A Bengali-first voice typing patch layer for Gboard 18.3.1. | Build instructions, static tests and source-only distribution notes. |
| SONU | A Tauri desktop voice typing app with local speech-to-text models. | Rust and React implementation, setup docs and release information. |
| Antigravity Claude Code Proxy | A local multi-provider gateway for Claude Code. | Provider adapters, configuration, test suite and security notes. |
| HoliBooks | A web reader for sacred texts from several religious traditions. | Reader implementation, data sources and local setup. |
The evaluation tools include offline or seeded examples. Those examples show the tools' behavior, not a measured ranking of live models or proof that an external service passes its requirements. Check each project's README for its current limits and verification scope.
- AI evaluation: repeatable tasks, withheld tests and inspectable results.
- API verification: OpenAPI contracts, requirements traceability and failure reporting.
- Software delivery: documented setup, test automation and clear operating limits.
- Voice interfaces: local speech-to-text and desktop tooling.
Start with its README, reproduce the documented example, then inspect the tests and output artifacts. CI badges report the checks configured by that project; they do not replace end-to-end testing in your environment.