Phase 5: Integration and evidence¶
Evaluate which parts of your proposed architecture have supporting evidence. A successful course lab demonstrates behavior in its own environment; it does not validate a different capstone deployment automatically.
Required deliverables¶
Complete Phase_5_Evaluation and Phase_6_Evidence in the portfolio. These are portfolio section identifiers; Phase_6_Evidence is an appendix, not the architecture defense presentation.
For each claim record: requirement, design decision, evidence source, environment, command or observation, result, limitation and next verification step. Label it:
- Demonstrated: implemented and tested within the stated environment.
- Simulated: behavior illustrated with a model or fabricated scenario.
- Proposed: design or control not yet implemented.
- Untested: implementation or claim lacks sufficient verification.
Use accessible text output, SQL, a small table, a textual lineage path or an annotated diagram. Screenshots are optional. Redact identities and secrets; submit privately.
Evidence categories¶
| Category | Relevant baseline | What to explain |
|---|---|---|
| Lifecycle | Labs 3–4 models and lineage | Dependencies and transformation behavior; a DAG alone does not prevent bypass. Identify additional enforcement needed. |
| Access | Lab 5 authenticated allow/deny checks | Actor, action, object and result; masking limits and untested roles. Generic lab results do not prove your own deployment. |
| Monitoring | Lab 6 client observations and simulated incidents | Label each source; discuss false positives, missed events and the absence of independent server audit collection. |
| AI governance | Lab 7 decision and lifecycle artifacts | Harm tradeoffs, transparency, accountable owner, review trigger and retirement implications. |
Example: "This DAG records dependencies. A separate permission check is needed to show that a reporting identity cannot bypass the intended access boundary."
Review and assessment¶
Complete peer review and revision before the defense. Assessment considers accurate evidence classification, reasoning, traceability, clarity and recognition of limits using the assessment descriptors. An honest proposed control with a sound verification plan is preferable to claiming that an untested control already works. Keep reference solutions and grading private.
Author: Isaac K. Nti.