I build software that tests, evaluates, and improves AI systems.
BS Computer Science · Four years of formal test operations in autonomous vehicles · Full-stack developer
My work sits at an unusual intersection: formal test discipline from safety-critical systems, project-based AI evaluation engineering, and full-stack SaaS development. I build things that work, break them to find out why, and then write down exactly what happened.
| Work | What it is |
|---|---|
| Together streaming evidence | Three synthetic SSE cases, pinned dependencies, deterministic replay, 8 passing tests. Delivered to the SDK's issue tracker. |
| Mastra token-estimation corpus | 16 cases / 64 estimates with reproducible checks; upstream maintainer closed the issue. |
| PrismML / Bonsai benchmarks | Multi-hardware benchmark rows with review corrections in public. |
| Hugging Face lifecycle validation | Independent before/after check of another contributor's fix. |
How I work: evidence-first; AI-assisted and I say so openly; no strings attached — if something I publish helps you, use it.
jayglidden.engineering@gmail.com github.com/EngineeringWithJay