Live in 3 tenants
Fifteen thousand
applicants.
One shortlist,
in days.
Awaaz is the hiring control room for high-volume recruiters. A four-stage AI funnel screens CVs, runs anti-cheat assessments, scores written work, and conducts a live voice interview with each finalist. At a variable cost of about PKR 8.59 per candidate.
- AHM-3401Ahmed R.→MCQ84
- FTM-2218Fatima K.→Written91
- USM-9087Usman A.→Voice78
12 sparkline · rolling
An entire hiring system,
built around the recruiter.
Pipelines, agents, automations, scoring rubrics, anti-cheat, billing. Six surfaces that behave like one product. No tabs hidden inside tabs. No "AI" sprinkled on top of a CRM.
Compose any funnel as a directed graph.
Drag stages onto the canvas. Pin rubrics, gates, and SLAs to each node. Versioned, with an audit log on every change, and dry-runnable on a sampled cohort before you commit.
Voice agents that conduct the interview, not read from one.
when stage = MCQ
and score ≥ 72
do advance, email,
schedule interview.
Trigger sequences in plain English or visual blocks. Idempotent retries, DLQ surfaced in the audit log, every send signed and traceable.
Rubrics as JSON. Versioned. Reviewable.
{
"stage": "MCQ",
"gate": { "score": >= 72 },
"weights": { "skill": 0.6, "lang": 0.4 }
}Catch the half percent who try.
Tab-switch, paste, secondary device, voice deepfake checks, copilot detection in written answers. Flags graded by severity, not boolean.
Every token, transcript second, and SMS, attributed to a candidate.
Tenant-scoped Redis emitter flushes to Postgres every minute. Billing knows what it's billing. Recruiters know what they spent.
Four users.
Four scenes.
One product.
The candidate sees a calm single column on 4G. The recruiter sees dense tables and a command palette. The tenant admin sees a quarterly summary. Ops sees a terminal deck. Same product, scene-driven chrome.
Processes 200 candidates a day. Hates clicks. Lives in keyboard shortcuts.
Operator deck / desktop
Dense tables · keyboard-first
- ›Virtualized tables, sortable to 50k rows
- ›Ctrl-K command palette for every primary action
- ›Side-by-side scorecard + transcript
- ›Bulk advance / reject with one keystroke
The trust ledger.
Q1 2026 / verifiedNumbers in this section come from production telemetry, not a brochure. They update on release. Every write in Awaaz is signed, every action is attributed to a user, and every impersonation event carries a reason and an auto-expiring scope. The audit log is read-only by design.
Plugs into the stack
you already trust.
First-class adapters for LLMs, voice, queues, vectors. Outbound webhooks signed with HMAC. Inbound automations through Zapier, n8n, Make, or a raw REST endpoint. If you have a CRM, it speaks to Awaaz in one config screen.
Push events to anywhere you want them.
- →Signed webhooks with replay protection
- →Event types: stage.advanced, score.computed, flag.raised
- →Per-tenant secrets, per-job filters
- →Dead-letter queue surfaced in audit log
// event · stage.advanced
{
"event": "stage.advanced",
"id": "evt_2k9f...",
"tenant": "sdh",
"candidate": "AHM-3401",
"from": "MCQ",
"to": "Written",
"score": 84,
"sig": "hmac-sha256:a3f9..."
}Trigger Awaaz from your own systems.
- ←REST API + typed TypeScript SDK
- ←Bulk candidate import, CSV or JSON
- ←Career-page form builder, embeddable anywhere
- ←Idempotency keys on every mutation
Real questions,
answered plainly.
No marketing prose. If the answer is "no" or "we're working on it", that is what you will see.
Awaaz screens a high-volume applicant pool through a four-stage AI funnel: CV screening with rubric + retrieval, MCQ assessment with anti-cheat, written assessment with plagiarism and copilot detection, and a live voice interview run by an LLM agent. It produces a final shortlist of high-confidence hires from ~15,000 applicants in roughly eleven days, at a variable cost of about PKR 8.59 per candidate.
We answer all of them ourselves.
No marketing inbox, no ticket queue. A founder reads every note. Average reply time is under three hours during PKT working day.