Every competency is recomputed from its own question scores, weighted by question. The model's evidence and summary are kept; its arithmetic is discarded. Unanswered questions count as zero.
The questions a security review asks, answered before you ask them.
Where the data sits, who can reach it, what the AI is allowed to decide, and what happens the day you leave. All of it is a design constraint rather than a setting.
Isolation, identity and residency.
Every business record carries a tenant, and access is checked in the database rather than the interface. A client cannot see another client. A vendor cannot see your pipeline.
Tenancy & access
- Row-level isolation on every business table
- Seven role levels, hierarchy enforced server-side
- Nobody can create or promote at or above their own level
- Client and vendor logins scoped to their own records
- Scoped API keys with per-resource field allow-lists
- Privileged action audit log
Accounts
- TOTP two-factor enrolment
- Device session list with per-device revocation
- Email on sign-in from a new device
- One-time codes that burn after five wrong attempts
- Admin-forced password reset
- Invitations bound to the role hierarchy
Data & residency
- Files, audio and video in your own bucket
- AWS Mumbai, Frankfurt, UAE Central, US-East
- Short-lived signed URLs, never public objects
- Consent captured before any recording begins
- Tenant backups and restore
- Full export to CSV or Excel at any time
What the AI is not allowed to decide.
An AI score a hiring manager cannot interrogate is worth nothing to them, and indefensible to a candidate who asks. These four rules are why the number holds up.
Scored on whether the answer addressed the question, was organised and was specific. Never fluency, grammar, accent or delivery — that would charge a transcription error against the candidate.
Pauses, tab switches and time away are reported to you and excluded from scoring by design. Each of those signals has innocent explanations more likely than the guilty one.
The recruiter confirms or overrides every outcome before a candidate moves stage, and the decision is written to stage history with the evidence attached.
Connect what you run, keep one record.
Google Meet, Teams and Zoom. Google and Outlook calendars. Slack, Teams and Telegram. Darwinbox and BambooHR. Zapier, Make and n8n — plus a scoped REST API and HMAC-signed webhooks when you would rather build it yourself.
Connect Your Stack
Job Boards & Sourcing
Publish openings across Naukri, LinkedIn, and Instahyre. Parse candidate resumes automatically and route directly into the correct mandate pipeline with source attribution intact.
// Webhook: candidate.sourced.v1
{
"event": "candidate.sourced",
"source": "naukri_premium",
"job_id": "job_devops_lead_08",
"candidate": {
"name": "Rohan Sharma",
"email": "rohan.sharma@example.com",
"experience_years": 5.5,
"current_ctc": "₹22,00,000",
"parsed_skills": ["Kubernetes", "ArgoCD", "AWS", "Terraform"]
}
}