The Jev System‑1 decision system.
This page openly explains the inner workings of the decision model used by client companies to prioritize their applications, calibrate eligibility probabilities and guarantee constant human oversight.
Decision matrix & criteria calibration
Expert TypeScript (6 years), Next.js 14 App Router, Server Components, Node.js, PostgreSQL, Drizzle ORM, Playwright, Jest, modular architecture.
Staff Engineer at a FinTech SaaS (3 years): rebuild on Next.js 14, LCP load time cut to 0.8s. Previously Senior Fullstack at Doctolib.
Paris (75). Valid EU work authorization. Availability: 1 month. Expectations: €72k gross. Hybrid accepted.
Nominal score with no eligibility penalty
Calibrated composite scoring
Rigorous weighting of the 5 decision dimensions rated between 0.0 and 4.0 by Jev (technical, substance of experience, authenticity of motivation, writing quality and soft skills).
As P_eligible approaches 0 (time-zone conflict, salary outside the range or no work permit), a discount of up to 80% is applied without arbitrarily removing the profile.
EU AI Act compliance (Annex III)
AI systems applied to recruitment and human resources management fall under the category of high-risk systems (Annex III). The Cevio platform is rigorously designed on the principle of decision assistance:
- No autonomous rejection: Jev orders and prioritizes profiles but never issues an outright rejection.
- Full auditability: Every recommendation comes with its JSON inference trace and the numerical justification of each criterion.
- Human accountability: Every shortlisted or rejected candidate is explicitly validated and logged by a qualified recruiter.