From JD to offer letter — and now the whole recruiting desk around it. Five accountable engines (ATS, CognitoChat, CognitoScreen, DeepTrust, CognitoView) plus a built-in CRM, semantic talent sourcing, workflow automation, talent pools & nurture, reporting, and an open API with signed webhooks. No bolt-ons. No third-party stitching. Every step accountable to a human reviewer at any moment.
Six capabilities that change what “ready for production” actually means. Each one written for a specific person in your buying committee.
Usage limits actually hold — not just report. No overage surprises at month-end. No retroactive invoice arguments. Plans cap themselves.
/health/ready and /services. Kubernetes probes that don’t lie. Status dashboards your ops team will actually trust.
Bill in INR, USD, EUR. Your customer doesn’t change processors. You don’t change codebases.
Routine cases never wait. Edge cases queue to a human with the critic’s full reasoning. Every escalation logged.
Every critical journey covered: CV upload, JD parse, bias check, billing rotation. 199 tests passing, zero failing.
One command — diagnose.py — tells you what’s wrong before ops has to ask.
The applicant tracking surface. Manages the full hiring lifecycle — sourcing, screening, interview coordination, offer management — without burying the reasoning.
Folder watch, email IMAP, manual upload. Unified pipeline. Deduplicated automatically. Tagged by source.
File hash · email match · phone match. Catches 15–20% wasted review work before it costs a single recruiter minute.
Scheduling, scorecards, stage tracking, SLA alerts. The recruiter’s calendar and the candidate’s never out of sync.
CV to ranked, scored, explained candidate in under a minute. End-to-end. No external API call.
Talk to your ATS in plain English. A real LangGraph supervisor decides. A critic LLM reviews every consequential action before it ships. Never silently confident.
The agent moves the candidate forward, sends the calendar invite, drafts the rejection — with the critic’s sign-off attached.
The case lands in your HITL queue with the critic’s full reasoning. Humans see what the AI was uncertain about, not just the result.
The chat window shows the supervisor’s reasoning, every tool call, every critic review. Auditable in real time.
phone-screen. Calendar invite drafted for Thu 3pm.768-dimensional semantic matching, not keyword filtering. Skills inferred from context. Experience calculated from actual dates. Evidence cited for every score.
"Backend engineer" matches "server-side developer". "Led a team" weighs differently from "team member". Context-aware.
Before a JD is published, the auditor flags exclusionary phrasing, gendered language, age-coded terms, and unrealistic combinations.
Incomplete profiles flagged. Ambiguous experience claims surfaced. Recruiters get signal, not just numbers.
Deepfake-resistant video interview verification. Synthetic profile detection. Credential cross-checking. The fraud filter every other ATS pretends doesn’t need to exist.
Face authenticity checks across the interview. Real-time alerts on artifacts. Recording with the critic’s reasoning attached.
Degree, employment, certification cross-checks. Flags inconsistencies in dates, claims, organisations.
1 in 6 applicants in remote roles fabricate credentials. DeepTrust catches them before the recruiter spends a single minute.
Bring your own model. Ollama, vLLM, OpenAI, Anthropic. DeepTrust’s core is platform-neutral.
Funnel velocity. Source quality. Skill-gap analysis. Bias reporting that satisfies legal. Predictive hiring success scores. Real-time dashboards your CHRO can defend in front of the board without a slide deck.
Where candidates stall. Where recruiters bottleneck. Where the data says the process is broken — not where you suspect it is.
EEOC-aligned breakdowns. Quarterly compliance reports auto-generated. The legal team finally has what they kept asking for.
Which channels deliver the candidates that make it past 90 days. Stop wasting budget on the channels that don’t.
Conversion: 0.92% · benchmark for SaaS engineering hires: 0.6%
The honest comparison. Where each approach actually lands on the things that decide a hire.
| Capability | Manual review | Keyword ATS | CognitoHire |
|---|---|---|---|
| Time to shortlist | 4–6 days | Hours, low quality | Under 60 seconds |
| Matching method | Human judgement | Keyword overlap | 768-dim semantic + evidence |
| Score transparency | Implicit, undocumented | Opaque ranking | Reasoning cited per score |
| Bias auditing | None | None | On every JD + outcome |
| Fraud / deepfake check | None | None | DeepTrust, built in |
| Human-in-the-loop | All manual | No structured review | Critic-gated, edge cases queue |
| Per-query AI cost | — | Scales with volume | Zero (local models) |
| Data sovereignty | On your desk | Vendor cloud | Self-hosted or managed |
| Audit trail | Scattered notes | Basic logs | Every decision, logged |
Keyword ATSs are faster than manual review. They are not more defensible. That’s the gap CognitoHire was built to close.
We deploy. We configure. You get defensible shortlists with full reasoning on your own real candidates. Live in under 24 hours. Zero commitment.