I built a surgical rehearsal tool in 2 weeks with no medical background. Here is what I learned from the first residents who tried it.
I am a CS person. I have never been in an OR. But I kept noticing a gap. Surgical residents spend nights studying procedures and still freeze when asked to perform them out loud under pressure. Reading about an incision and making one are completely different things. Nobody trains for that gap.
So I built OR Ready. Residents speak through each step of a procedure, get instant scored feedback, and see it unfold in 3D anatomy. No headset, just a browser. Two weeks in. Here is what I have learned so far:
1. The login friction killed my first pilot. My first tester could not get past the auth screen for two days. Nothing in the code told me that. She did.
2. Residents are genuinely interested but genuinely busy. 80 hour weeks are real. Getting feedback out of someone between night shifts takes patience not pressure.
3. The medical content has to be exact. One wrong term in the evaluator and the whole thing feels untrustworthy. I brought in an MBBS reviewer and it changed everything.
4. The product works best when I get out of the way. The moment a resident says "I actually missed that safety check" is the moment I know the thing is real.
First module is live at getorready.com. Distal radius volar plate fixation, 12 steps, scored feedback on every attempt.
What would you do differently in week 3?
The part that jumped out at me was bringing in an MBBS reviewer.
A lot of products can survive being slightly wrong.
This feels like one of those categories where users need to trust the feedback before they can benefit from it.
I'd be curious how you're thinking about validation there, because "the simulation works" and "the resident trusts the simulation" aren't necessarily the same milestone.
100%. That is one of the biggest problems I’m trying to solve right now.
For the first procedure, I worked with an orthopedic surgeon friend to build the clinical script. We used standard teaching material as the base, then tried to make the flow closer to how a resident might actually talk through the case in the OR. The goal was not just “what are the textbook steps?” but “what should the resident remember in the OR?”
For evaluation, I’m being careful not to let a general LLM freely judge the answer. The scoring is based on a structured rubric with required concepts, acceptable wording variations, and safety-critical checkpoints. So the user does not need to say the exact sentence, but they do need to hit the important clinical ideas.
And yes, trust is the core issue. A tool like this is only useful if the resident believes the feedback is clinically reasonable. Right now my validation process is basically:
Clinician review of the script and rubric
Residents trying the module
Tracking where the evaluator feels too strict, too easy, or unclear
Adjusting the rubric with clinical input, not just based on my own assumptions
That is also why the pilot feedback matters so much. I’m not just asking, “Did you like the product?” I’m trying to find out whether the evaluator feels fair, whether the clinical flow feels real, and whether residents would trust it enough to use before a case.
That's the distinction I was wondering about.
A resident can tell you the evaluator feels fair, realistic, and clinically sound.
The harder question is whether those reactions are actually validating the simulation itself or validating something adjacent to it.
Those can end up being very different conclusions.
Happy to share the fuller thought if useful. What's the best email to reach you on?
Hey, I'd love to hear your thoughts on it. Please reach out to me at anirudhparakala@gmail(not the full email because I'm not allowed to post links yet)