Three tests. One honest snapshot.
A three-engine online assessment platform — IQ, Psychometric (OCEAN + RIASEC), and General Knowledge — for students, career counsellors and corporate L&D. Claude writes the narratives, Razorpay handles UPI, AWS issues a QR-verifiable certificate. Editorial in tone, honest by design.
- IQ + Psychometric + GK · separate engines
- Claude-written narratives · no jargon
- QR-verifiable PDF certificates
- UPI · cards · coupons · free mode
Built on infrastructure recruiters and institutions already trust
An assessment platform with editorial register and engineering integrity.
Most online IQ tests are loud, click-bait, and intellectually dishonest — "Your IQ just changed!" doesn't happen, ever. CognitiveIQ treats the reader as an adult. Three real engines, AI-written prose that doesn't overclaim, certificates that recruiters can actually verify, and a Postgres / Drizzle / Claude stack engineered to scale.
Honest by design
We say "test performance improves," not "your IQ goes up." We say "Emotional Sensitivity," not "Neuroticism." Personality results are framed as "a snapshot, not a verdict." The narrative generator is hard-coded against the words amazing, impressive and perfect — and against exclamation marks.
No question ever twice
Every served question is logged to user_question_history. The next session joins it out before sampling. Combined with ~5,000 IQ + ~5,000 GK items, retake fatigue is not a problem. The pgvector ivfflat index also dedupes the bank itself — two questions that mean the same thing don't both make it in.
Certificates that verify
16-character IDs (CIQ + 13 from a no-ambiguity alphabet), private S3 storage, QR code on the PDF that resolves to a rate-limited public verify page. A recruiter scans, sees the candidate's name and band, and trusts the result — no login required.
Four layers. 24 modules. One data model.
The three test engines, the AI content layer, the certificate & verification rail, and the payments / auth / admin layer — sharing one Postgres schema and one audit log.
What you can do · who you are · what you know.
Three independent scoring pipelines. Each can be taken on its own, or bundled. None pretends to do the other's job.
- IQ testPattern recognition · analogies · spatial · numerical · verbal. Weighted-raw → base IQ → age-adjusted final. ~5,000-item bank, no-repeat across sessions.
- Psychometric (OCEAN + RIASEC)120 IPIP-NEO items measuring the Big Five (re-labelled Emotional Sensitivity) + 30 RIASEC items mapping career interests. Likert, no right/wrong.
- General Knowledge50 items across current affairs, science and civics, scored as percentage with 6 bands. ~5,000-item bank, balanced by sub-topic.
Certificate pipeline
Triggered on test submit, fully idempotent. The candidate's name, photo (if uploaded), test type, score band and date are baked into a PDF rendered with @react-pdf/renderer in the editorial type system — Fraunces display + Figtree body — then pushed to private S3. A QR code on the certificate resolves to the public verify page.
- 16-char ID — CIQ + 13 from 29-char alphabet
- @react-pdf/renderer · Fraunces + Figtree fonts
- Private S3 · presigned download on demand
- Idempotent — same session, never two certs
- Optional Rekognition-verified photo on PDF
Public verification
A recruiter scans the QR. The /verify/{certNumber} page opens — no login, no app install, no friction. The page shows candidate name, test type, score band, issue date. Rate-limited at 60 / minute / IP so enumeration attacks fail. The candidate's underlying answers are never exposed.
- Public — no login required
- Rate-limited 60 / min / IP
- QR resolves to /verify/{certNumber}
- Candidate name + band + date only
- Underlying answers never exposed
Editorial prose. Personal. Honest. No exclamation marks.
Every submitted session gets a 3-paragraph result narrative written by Claude Sonnet — using the candidate's first name once, naturally, in editorial register. No bullets. No headings. No "amazing" or "impressive." Cached in the database so each session triggers exactly one model call.
- Score in age-band context
- Per-domain breakdown in prose
- Concrete strengths + growth edges
- "Test performance improves" framing
- Never says "your IQ went up"
- 3 paragraphs, no bullets
- OCEAN profile in prose
- RIASEC career-cluster suggestion
- "Snapshot, not a verdict" framing
- Never uses "neuroticism"
- Emotional Sensitivity instead
- Suggests next step, not a label
Public SaaS · branded for your institution · on-prem.
Same engine, same certificates, three operating models — pick the one that fits your institution and budget.
Public SaaS
Hosted at cognitiveiq.in — candidates register, pay via UPI, take the test, get the certificate. Free-mode windows + coupons for campaigns.
Branded for your institution
Custom domain, your logo on the certificate, cohort-coupon import, admin reporting across the group. Schools, colleges, coaching institutes.
On-prem / air-gapped
Single docker-compose deploy on a server you own. SES → SMTP swap, AWS Rekognition → manual moderation. University IT departments.
Plays nicely with the assessment ecosystem.
Razorpay for Indian payments, AWS for the heavy media lifting (Rekognition + SES + S3), Anthropic for the brain. Redis for rate limiting and BullMQ. Every integration has a clear failure mode — Redis fails open, SES dev-mode logs to console, Razorpay webhook is HMAC-verified.
Eight first-party integrations out of the box. Custom integrations — SIS / LMS bridge, Stripe for non-INR markets, on-prem mail relay — are fixed-scope.
Locked, pinned, engineering-grade.
Next.js 16.2 + React 19 + TypeScript 5.7 strict. Drizzle on self-hosted Postgres 18 with pgvector. Tailwind v4 CSS-first @theme. bcryptjs + SHA-256 sessions. AWS for S3 / Rekognition / SES. Anthropic Claude for the AI work. Docker compose for deployment.
- Next.js 16.2 App Router + Turbopack
- React 19 + TypeScript 5.7 strict
- Tailwind v4 (CSS-first @theme)
- Fraunces (display) + Figtree (body)
- Editorial design — no stock photos
- RSC by default; small client islands
- Drizzle ORM (no Prisma)
- PostgreSQL 18.3 + pgvector 0.8
- Self-hosted DB (not RDS)
- Redis 7 for BullMQ + rate limit
- bcryptjs cost-12 + SHA-256 sessions
- Zod-validated env + API inputs
- Anthropic Claude Sonnet 4.6
- Question generator + narratives
- pgvector ivfflat dedupe
- AWS Rekognition (photo)
- @react-pdf/renderer (PDF)
- nanoid (certificate IDs)
- Docker + docker-compose
- AWS EC2 + S3 + SES (production)
- 90-min runbook from zero to live
- AWS_DEPLOYMENT_RUNBOOK in repo
- pm2 reload zero-downtime deploys
- Migrations idempotent + tracked
Anywhere a real assessment is needed.
K-12 diagnostics to university career counselling to corporate graduate hiring. Same engine, configured for the cohort and the question.
90 minutes for SaaS · 1–2 weeks for branded.
The AWS_DEPLOYMENT_RUNBOOK in the repo is the script — 90 minutes from a fresh AWS account to a working URL. For a branded institutional rollout we layer domain, certificate template, cohort import and admin training on top.
Fraunces + Figtree, deep-navy / warm-ivory palette, magazine-style eyebrow text — the candidate experience reads like a journal, not a quiz site.
Discovery
Use-case scoping — schools / corporate / public SaaS — cohort size, test mix, branding requirements, hardware constraints.
Provision
AWS or on-prem servers up, Postgres 18 + pgvector + Redis, domain + TLS + SES verified, Razorpay account live.
Bank seed
Load IQ + GK + psychometric banks, admin-review the AI-generated batch, set test parameters and bands.
Brand + cohort
Customise certificate template + logo + footer; bulk-import cohort users / cohort coupons / free-mode windows.
Go-live
Smoke-test end-to-end (register → pay → test → cert → verify), enable webhooks, switch DNS, open to candidates.
Hypercare
4 weeks of daily standups + admin training, then SLA-based managed support + quarterly review.
Real SLAs, calibrated to an exam window.
Two assessment stories — placeholders till the real ones land.
Real institution names and metrics available on a discovery call.
Outsourcing OCEAN + RIASEC testing to a foreign vendor at ₹400 / student. Results took 7 days to come back. The vendor used "Neuroticism" as a label and parents complained. No verifiable certificate — students used screenshots.
Branded CognitiveIQ deployment with the firm's logo on the certificate, custom domain, cohort coupons. Switched language from Neuroticism to Emotional Sensitivity, added Claude-written 3-paragraph narratives. End-to-end student experience under 30 minutes.
- Result delivery: 7 days → instant
- Cost per student: ₹400 → ₹110
- Parent CSAT (language): 64 → 91
Paper-based annual diagnostic took two weeks to score; results never reached individual students; the placement cell had no data when it was time to map remedial training.
On-prem CognitiveIQ deployment in the college server room. Cohort import via CSV, free-mode active for the diagnostic, certificates auto-emailed. Placement cell gets aggregate dashboard the same evening.
- Score-to-student turnaround: 14d → same evening
- Placement cell remedial mapping: lag → real-time
- Annual cost vs paper + manual scoring: −68%
Common questions
Three separate engines for three separate questions. The IQ test measures fluid reasoning across pattern recognition, analogies, spatial rotation, numerical and verbal items — scored as a 3-step weighted-raw → base IQ → age-adjusted final. The Psychometric test is a 120-item IPIP-NEO (OCEAN: Openness, Conscientiousness, Extraversion, Agreeableness, Emotional Sensitivity) plus a 30-item RIASEC career-interest profile — Likert-scored, no right/wrong. The General Knowledge test is a 50-item current-affairs / science / civics test, scored as percentage across six bands. A learner can take any one, two or all three.
30-minute demo with your test mix and your branding.
Tell us your institution type (school / college / career firm / corporate), test mix (IQ / Psych / GK / bundle) and cohort size. We'll demo with a configured sandbox using your real branding — not a generic vendor reel.
- 30-minute discovery + tailored demo
- Live test-taking + certificate walkthrough
- Branding + cohort-import deep-dive
- Deployment plan (SaaS / branded / on-prem)