How Surgri Harnesses AI to Continuously Expand and Refine Plastic-Surgery Questions
Surgri began as a simple idea: can generative AI help give plastic-surgery trainees a smarter, faster and cheaper way to cover the curriculum. One month later we have a living question bank that grows every day—thanks to a unique human-in-the-loop AI workflow, all running on our lightweight Rails 8 and Tailwind CSS stack.
A Rails-Powered Foundation
Surgri is built with Rails 8, deployed on Heroku, and organised around a flexible data model:
- Questionable reverse-polymorphic association lets each question belong to any format—MCQ, extended-matching, image-based viva and more.
- Categories map directly to the Intercollegiate Surgical Curriculum chapters, so you can target exactly what you need.
- Stacks save your chosen categories, while blocks assign fresh questions into each practice session.
This foundation means we can swap in new question types—or new AI models—without painful refactors.
The Three-Step AI Creation Cycle
| Step | What Happens |
|---|---|
| 1. Generate | A large-language model analyses the official curriculum and drafts new questions, answer options and explanations. |
| 2. Curate | Our editorial team of credentialed surgeons reviews each draft. They can edit, accept or reject with a single click. |
| 3. Refine | Accepted questions are re-analysed by AI, incorporating editor changes plus anonymised user feedback and performance data (difficulty, discrimination, comments). The AI adjusts wording, distractors and explanations to boost clarity and educational value. |
The result: every question you see is both human-verified and data-driven.
Why This Matters for Learners
- Curriculum Coverage, Fast – We can spin up hundreds of fresh questions whenever guidelines change.
- Quality You Can Trust – Human editors catch nuance and clinical edge-cases that AI alone might miss.
- Constant Improvement – Your attempts and comments feed right back into the cycle, raising the quality bar week by week.
What’s Next?
User feedback questionnaires will be given out at various milestones of usage. The more you rate questions and comment, the better the questions get.