Overview
An embassy interview lasts a few minutes and can decide years of someone's life. Intertricks was built to make preparing for it structured, practiced and measured.
Applicants had been preparing the way people prepare for anything terrifying and unfamiliar, asking friends who'd been through it, watching videos of strangers, guessing at what confident is supposed to sound like. Intertricks came to us with a clear thesis, this preparation could be structured, practiced and measured. Applicants could rehearse real interview questions, by text and on video, and receive a scored assessment of how they did, while the consultancies who already advise these applicants could offer the same preparation as part of their own service.
Challenge
The AI had to be useful without being trusted blindly, since an applicant wrongly told they're ready walks into an embassy over-confident, so any AI assessment needed an honest route to human judgement alongside it. Video answers are heavy too, they need to be captured, stored, transcribed and scored reliably on ordinary consumer internet connections, without the applicant waiting around wondering whether it worked. Two audiences shared one platform, individual applicants wanting practice and progress, agencies wanting oversight across their whole roster, a product design problem before it was an engineering one. And results had to be verifiable, a score is worth little if anyone can screenshot and edit it, so verification needed to be built into the artefact itself.
Approach
Transcription and scoring were built as separate stages, video answers are transcribed first and the resulting text is what gets scored, which makes the pipeline debuggable, you can see exactly what the model heard before questioning what it concluded. Human review was built as a first-class feature rather than a complaint channel, applicants can request a human review of the same attempt, one click away rather than bolted on as a support ticket, which is the design decision that keeps the AI honest. Applicants and agencies were given separate surfaces for separate jobs on the same underlying platform and data model, practice and progress for applicants, a roster-based dashboard for agencies to register, assign and monitor students. And every result carries a QR code that resolves to a live verification page, so a consultancy can scan it and confirm the result is real, turning a score into a credential rather than a screenshot.
Result
Intertricks is a live platform serving both individual applicants and registered agencies, with AI-graded assessments and a human review gate actually shipped in production, a pattern more often described than built. Specific figures for users, completed attempts and registered agencies are available on request.
