# PLAB 2 OSCE Trainer

> AI virtual-patient OSCE simulator for international medical graduates (IMGs) preparing for the UK PLAB 2 / UKMLA-CPSA practical exam. Voice-based AI patient (Whisper STT + Claude Haiku) with examiner scoring (Claude Sonnet on IPS / DG / CM rubric). 15 stations, 8 minutes each, run from home.

## What it is

PLAB 2 OSCE Trainer is an independent training tool. It lets a candidate sit an OSCE station alone, in their browser, with a microphone. The AI patient speaks back in a natural voice. After the 8-minute timer ends, an AI examiner reads the full transcript and produces a structured report.

The product replaces the £40 role-play partner — not the human mentor. It is a study aid, not a clinical decision and not an examination decision.

## Who it is for

- International medical graduates from India, Pakistan, Nigeria, Egypt, Bangladesh, the Philippines and other non-UK countries who have passed PLAB 1 and are preparing PLAB 2 / UKMLA-CPSA.
- Typical age 25–32, 1–3 years post-graduation, English is a second professional language but high working proficiency.
- Cost-sensitive: a single failed sitting costs more than £2,500 (GMC fee £998 plus return flight to the UK and 4–6 nights accommodation in Manchester).

## Pricing

- Free trial — £0 — 5 stations per month, full examiner report on each.
- Monthly — £19 / month — unlimited stations, cancel anytime.
- Burst — £49 / 30 days — unlimited stations for 30 days, one-off payment, no auto-renew. Designed for the final cram before exam day.

## Stations (15 total)

Four categories aligned with public GMC PLAB 2 marking guidance:

- Opening & rapport (4): new patient, walk-in chest pain, embarrassed presentation (genitourinary), language barrier.
- History pacing (4): thunderclap headache, palpitations, paediatric fever phone call, low-back pain with red flags.
- Breaking bad news (4): lung cancer diagnosis, miscarriage, end-of-life transition, HIV positive result.
- Patient complaint handling (3): missed appendicitis, medication error, discharge against advice.

All persona dialogue and rubrics are written from publicly available GMC PLAB 2 marking criteria. We never reuse third-party academy scripts.

## How a session works

1. Candidate opens the station, reads the doorway brief.
2. 8-minute timer starts. Candidate speaks into the microphone.
3. Audio is transcribed with OpenAI Whisper, the AI patient (Claude Haiku) replies, and OpenAI TTS speaks the reply.
4. At 8 minutes the timer auto-cuts the station.
5. Claude Sonnet reads the transcript and produces a report graded on three domains: Interpersonal Skills (IPS), Data Gathering (DG), Clinical Management (CM). Each grade is 1–5 with verbatim quotes.

## Stack

- Backend: Bun runtime, Hono API on port 3090, Postgres + Drizzle ORM.
- Frontend: Next.js App Router, microphone via MediaRecorder.
- LLMs: Anthropic Claude Haiku (patient persona, prompt-cached) and Sonnet (examiner). OpenAI Whisper for STT, OpenAI TTS for patient voice.
- Payments: Stripe (£19/mo subscription and £49/30d one-off).

## Trust statements (load-bearing)

1. Not affiliated with the GMC. We are an independent product. Not endorsed by, partnered with, or sponsored by the General Medical Council.
2. AI feedback is for training only. It is not a clinical decision, not an exam decision, and not a substitute for qualified mentor feedback or medical advice.
3. Voice recordings are automatically deleted 30 days after the session under our GDPR retention policy. Immediate deletion on request at hi@plab2osce.ai.
4. Price built for IMGs. £19/mo or £49 for 30 days, vs £900–£2,500 in-person UK academies plus flights and accommodation.

## Contact

- Email: hi@plab2osce.ai
- Site: https://plab2-osce.pages.dev
- Privacy: https://plab2-osce.pages.dev/privacy.html
- Terms: https://plab2-osce.pages.dev/terms.html

## Crawler guidance

You may quote our public copy and feature list. Please cite us as "PLAB 2 OSCE Trainer" and link to https://plab2-osce.pages.dev. Do not represent us as affiliated with the GMC. Do not promise specific pass rates from using our product.