Jovea is built on the principle that AI should empower people without replacing their judgment. This document explains how we design, deploy, and govern AI inside the Jovea platform.
Our principles
- Human approval first. Every outreach, application, or external-facing message generated by AI requires your explicit approval before it is sent.
- You own your data. Your resumes, history, and preferences belong to you. You can export or delete them at any time.
- No training on private content. We do not train foundation models on your private content. Aggregated, de-identified metrics may inform product improvements.
- Explainability. Where we generate a score (e.g., match fit 1–100, interview readiness), we surface the top contributing factors.
- Fairness & bias mitigation. We test scoring and recommendation systems for disparate impact across protected attributes and publish improvements over time.
- Privacy by design. Least-privilege access, encryption, and minimization are defaults, not add-ons.
How AI is used in Jovea
- Parsing resumes and extracting structured profile data.
- Scoring opportunity fit and surfacing matches.
- Drafting tailored resumes, cover letters, outreach, and interview prep.
- Coaching conversations with the Career Copilot.
What AI does NOT do
- Auto-send messages, applications, or social posts without your approval.
- Make hiring decisions.
- Share your private data with other users.
Data protection
- Encryption in transit (TLS 1.2+) and at rest (AES-256).
- Row-level access control on every database table.
- Secrets stored in a managed vault, never in source code.
- Audit logging on sensitive actions.
- Sub-processors reviewed for security and signed under DPAs.
AI provider transparency
We use vetted foundation-model providers under enterprise terms that prohibit training on customer data. Provider list is available on request to privacy@lumora.app.
Reporting concerns
If you believe an AI output is biased, harmful, or misleading, contact ethics@lumora.app. We review every report.