This portfolio documents cases where AI-assisted workflows were used to reproduce, test, audit, or clarify technical claims across domains.
This portfolio is evidence of AI-orchestration skill: the ability to enter unfamiliar technical domains, build source-grounded workflows, direct local tests, preserve uncertainty, and produce artifacts that domain experts or maintainers can engage with.
Public repository: bankszach/ai-orchestrated-technical-audits
- Portfolio index: PORTFOLIO_INDEX.md
- First case executive summary: case_studies/001-eml-qualified-reproduction-audit/executive_summary.md
- Responsible representation policy: docs/representation_policy.md
- Evidence levels: docs/evidence_levels.md
This portfolio is designed to show AI-operator skill through reproducible technical artifacts. The emphasis is not on claiming native expertise in every audited domain, but on demonstrating the ability to structure investigations, use AI systems responsibly, build local tests, document uncertainty, and communicate useful findings to experts or maintainers.
- AI-orchestrated research and testing portfolio.
- Reproducibility and technical audit work.
- Evidence of workflow design and operator skill.
- Domain-agnostic contribution record.
- Not claiming domain expertise in every field.
- Not a collection of AI-generated essays.
- Not claiming every audit result is final truth.
- Not claiming expert approval unless explicitly documented.
- Not quoting private correspondence without permission.
- Not using AI output as a substitute for reproducible artifacts.
- Not a replacement for formal peer review.
The first portfolio case is an AI-orchestrated qualified reproduction audit of the EML single-operator elementary-functions paper. The public artifact is a GitHub repository with tests, reports, and a status matrix. The audit identified a branch-semantics issue in the inverse-function discovery/verification chain. The paper author responded in private correspondence and acknowledged the specific issue; that correspondence is not quoted publicly and is not represented as approval of the full repository.
Domain: symbolic computation / complex elementary functions
Public artifact: eml-qualified-reproduction-audit
Case study: case_studies/001-eml-qualified-reproduction-audit/
Executive summary: case_studies/001-eml-qualified-reproduction-audit/executive_summary.md
The evidence model is documented in docs/evidence_levels.md.
External interactions are governed by docs/representation_policy.md.
- PORTFOLIO_INDEX.md - case list and evidence summary.
- case_studies/001-eml-qualified-reproduction-audit/ - completed EML portfolio case.
- docs/ - methodology, evidence levels, representation policy, and operator skill map.
- backlog/ - placeholders for future candidate tracking.
- scripts/ - local portfolio validation checks.
The repeatable audit methodology is documented in docs/audit_methodology.md.
Future paper candidates are tracked in backlog/candidate_papers.md.
Future repository candidates are tracked in backlog/candidate_repos.md.
Zach Banks
GitHub: bankszach