Public score card · scored May 5, 2026
MACH Score Card
Architecture review for AI feature in fintech app
Anonymous candidate
Total
41
of 100
The mission requires a written critique (max 1500 words) and a risk matrix for an AI-powered transaction categorization feature. The candidate must identify three risks (technical, regulatory, UX) with proposed mitigations. EVIDENCE OF GAPS: 1. **Deliverable completeness**: The write-up is approximately 70 words, far below the 1500-word maximum. While brevity can be valuable, the mission explicitly asks for a 'writt
The submission's write-up is a single 79-word run-on sentence lacking any structural elements (no sections, headings, or formatting). The mission explicitly requires a 'written critique (max 1500 words)' and a 'risk matrix' as separate deliverables. This submission provides neither a proper critique document nor a risk matrix—it's essentially abbreviated notes compressed into one paragraph. WEAKNESSES: (1) Complete a
STEP 1 – Evidence gathering: The write-up identifies three risks (model hallucination, GDPR, silent reclassification) with corresponding mitigations (confidence threshold + fallback, on-prem/anonymization, visible badges + retraining). Process notes mention reading design twice, drafting per stakeholder lens, using Claude for stress-testing, and cross-referencing PSD2/AML. Time reported: 90 minutes for a 4-hour missi
First, examining weaknesses with evidence: (1) The write-up is severely truncated at ~90 words when the deliverable specified 'max 1500 words.' While brevity can be valuable, this submission reads like bullet points or an executive summary rather than a 'written critique.' It lacks substantive explanation of *why* each risk matters, *how* mitigations would be implemented, or *what trade-offs exist*. For example, 'on-
First, identifying concrete weaknesses with evidence: 1. **Surface-level risk identification without depth**: The write-up lists three conventional risks that any competent reviewer would identify through a basic checklist approach. 'Model hallucination,' 'GDPR concerns with LLM providers,' and 'silent reclassification eroding trust' are standard concerns for any AI feature in fintech. The submission doesn't demonst
This page shows shareable evaluation metadata only. Mission titles may appear as "Confidential mission" when the hiring company keeps the role private.