This portfolio validates AI-orchestration skill, not necessarily deep native expertise in every target domain.
| Skill | Portfolio Evidence |
|---|---|
| Problem framing | Selects bounded claims that can be tested or clarified. |
| Agent orchestration | Directs AI systems across source review, test design, writing, and verification. |
| Source acquisition | Gathers papers, repositories, reports, and implementation context. |
| Literature triage | Identifies the parts of a paper or repo that matter for a reproducible claim. |
| Test design | Converts claims into local checks, fixtures, witnesses, and regression tests. |
| Local reproducibility | Builds runnable artifacts and records pass/fail status. |
| Codebase scaffolding | Structures repos so reports, tests, and evidence are inspectable. |
| Report writing | Produces restrained summaries that separate result from inference. |
| Uncertainty management | Preserves ambiguity when a result is promising but source-ambiguous or not verified. |
| Expert communication | Contacts authors or maintainers with narrow, reviewable findings. |
| Domain transfer | Applies a repeatable audit method across unfamiliar technical areas. |
The transferable skill is operating AI systems and local computation to make technical claims more inspectable, testable, and useful to domain experts.