Talk to us
◆ Miraai · Your future, guided
Most people decide who to become with a quiz from another century and advice from whoever's nearby. Miraai replaces that with a guided journey — a cinematic assessment, an AI copilot, and live maps of every job and pathway — built on open, published science.
◆ 01 · The problem
Over a billion young people are choosing what to study, where to work, and who to become — and the systems meant to guide them weren't built for any of it. The cost shows up everywhere: in NEET rates, in skills mismatch, in talent that never finds its place.
Counselor ratios are unmanageable, and the students with the fewest professional networks at home get the least support at school. The advantage gap compounds: exposure, internships, and mentors flow to those who already have them.
Static interest inventories and type quizzes ignore values, context, and emotional readiness — and they're trivially easy to second-guess. They tell people what box they fit, not who they could become or what to watch out for on the way.
Recommendations rarely touch the live labor market — what's growing, what's declining, what it actually takes to get in. Misalignment between education, values, and real pathways costs economies billions and individuals their momentum.
The fix isn't a better quiz. It's a better journey — from self-understanding to a concrete, realistic plan.
◆ 02 · The journey
Miraai is one connected journey in four movements. Each one feeds the next, and an AI copilot carries the thread — so insight turns into direction, and direction turns into a plan.
Discover
A cinematic, image-based experience that feels like a story and reads like science. Forced-choice scenarios surface instinctive preferences — personality, interests, values, and abilities — including where a strength tips into a risk. No grids of statements. No box at the end.
Understand
One continuous, natural conversation — no modules, no menus. Mira explains results, asks the questions a great counselor would, and nudges follow-through. Chats disappear by design; what's remembered is structured and minimal — goals, results, next steps. Personalization without surveillance, with safeguarding built in. Mirror, map, mentor.
Explore
A continuously refreshed map of real labor-market demand, matched on psychological fit — traits, values, environment — not keyword overlap. Every match explains why it fits, what the growth outlook is, and which doors it opens next.
Build
Multi-route blueprints from here to there: degrees, certifications, apprenticeships, and stackable alternatives — filtered by cost, geography, duration, and accessibility, with low-cost and high-equity options surfaced first. Several honest ways in, not one gilded one.
The same architecture powers early-careers programs, internal mobility, and national employability initiatives — guidance that doesn't end at the score.
◆ 03 · The science
Engagement isn't decoration — it's measurement strategy. People answer honestly when an experience feels instinctive rather than interrogative. Here's how we changed the way assessment and guidance work together.
Scenario imagery puts you inside the moment instead of in front of a statement. It lowers the literacy barrier, travels across cultures, and surfaces gut-level preference — which is exactly the signal self-reports polish away.
Forced-choice trade-offs and adaptive testing mean there's no obvious "right" answer to perform — even with an AI assistant open in the next tab. We benchmark our formats against LLM-assisted faking directly, as a published study.
Real behaviour is curvilinear: confidence tips into arrogance, diligence into rigidity. Our adaptive scales see where a strength becomes a risk — the difference between a label and genuinely useful guidance.
The Miraai Open Research Program takes the assessment from psychometric foundation to deployment-grade evidence. Every study is pre-registered on OSF before data collection; materials, analysis code, and de-identified data are released with the results. Select any study to see what it proves.
Establishes the instrument's factor structure, reliability, and calibration — and settles a question nobody else is asking in public: does image-based delivery measure the same thing as text? Every downstream claim is anchored here.
Can LLM personas stand in for human pilot samples? 1,000 richly conditioned personas across three frontier models, calibrated against our human benchmark and tested on held-out items and held-out people. The result: a validated synthetic-respondent pipeline that pre-screens items before a single participant is recruited — faster development, without lowering the evidential bar.
Likert, single-stimulus, and forced-choice formats go head-to-head under honest, fake-good, and LLM-assisted faking conditions. If an assessment can't survive a candidate with a chatbot, this is where it shows — in print.
Does the instrument mean the same thing in six culturally distinct samples? And does first-person imagery travel better than text? Measurement invariance, tested properly — with synthetic respondents pre-screening item bias before expensive international recruitment.
The closing loop: do our adaptive scales predict counterproductive behaviour and leadership derailment beyond what linear trait scores explain? Real organizational outcomes — the evidence no synthetic method can substitute for.
This program is an open invitation. Where each study would benefit from external collaboration, we say so explicitly — and we welcome labs, partners, and reviewers at any stage.
Collaborate on a study◆ 04 · The Majdal engine
Project Majdal is Miraai's AI-native psychometric pipeline. It takes a definition of what you want to measure and produces a screened, bias-audited, statistically pre-tested item pool the same day — with a complete audit trail on every single item, and a qualified psychologist making the final call. What used to take months of drafting and committee review now takes hours.
Construct definitions become observable, real-world behaviours — each validated for construct, face, and discriminant validity before a single item is written.
Hundreds of candidate items, generated against tight design briefs and anchored to specific behaviours — not vague restatements of the trait.
Every item screened for reading level and rated by multiple independent AI reviewers across six fairness dimensions — gender, ethnicity, age, religion and culture, socioeconomic status, orientation and disability.
Five independent models act as subject-matter experts, rating how well each item measures its target — and only its target. Items clear strict thresholds or they're out.
Language embeddings simulate how items will behave statistically before any human data is collected — flagging weak items while changes are still cheap.
A six-factor personality model enriched with adaptive scales that distinguish a strength from its shadow — confident vs. overconfident, decisive vs. impulsive. Delivered as immersive, image-based forced-choice pairs that surface instinctive preference rather than rehearsed self-presentation.
Under the hood: modern forced-choice measurement (Thurstonian IRT, GGUM, MUPP) — the faking-resistant kind. Full detail in our published protocols, if you like that sort of thing.
The pipeline replaces the drafting and screening — never empirical validation. Every automated decision is reviewable and reversible; a psychologist decides what survives. The architecture has been independently reviewed by Dr. Philseok Lee of George Mason University.
Every item carries full data lineage: concept → item → audit → score.
◆ 05 · SJT studio
The same production capability powers our situational judgment tests. We build bespoke SJTs from your organization's real critical incidents — then our AI production line re-skins validated video scenarios for any industry, region, or cast. Same decision point, same response options, same scoring. New world. Days, not quarters.
Scenarios come from structured critical-incident work with your SMEs and are mapped to your competency framework — so the judgment being tested is the judgment the job actually demands.
Our enhanced scoring weights penalties by expert agreement: deviate from a unanimous expert key and it costs you; hedge to the midpoint and it no longer pays. Sharper distributions, more defensible decisions.
Characters are extracted into canonical references and locked across every scene — no visual drift, no accidental cues, cinematic quality, neutral professional narration. Assessment integrity is a rendering rule, not an afterthought.
What never changes: decision points, response options, effectiveness keys. That's the validity you're paying for — re-skinning protects it by design.
◆ 06 · The people
Co-Founder & Chief Product & Innovation Officer
Organizational psychologist and talent assessment specialist with a career spanning Asia, Europe, the Americas, and the Middle East. Former senior roles at HireVue, Psytech International, and Podium Systems. Architect of Project Majdal — Miraai's scientific and assessment backbone.
Co-Founder & Chief Executive Officer
Chartered Psychologist and Associate Fellow of the British Psychological Society. Former CEO of Podium365 (online psychometrics) and five years evaluating ML outcomes at Microsoft Bing. Leads Miraai's strategic direction, operations, and global positioning.
Co-Founder & Chief Revenue Officer
Senior organisational psychologist with nearly 30 years of experience in recruitment and talent systems. Former APAC Consulting Director for a major test publisher. Directed the New Zealand Government Scholarship programme serving 9,000+ applicants from 97 countries. Leads Miraai's commercial strategy and partnerships.
◆ 07 · Next step
Twenty minutes with the assessment on a phone says more than any deck. If you'd rather interrogate the method first — good. That's the kind of partner we built this for.
