Chapter 06 / Applied AI · Education
MedAI.
Structured feedback for medical presentation practice.
The idea
Turn a practice transcript into sections, reasoning, and rubric-based feedback.
Python / FastAPI / React / S3 / MongoDB
Animated architecture
A structured evaluation pipeline
This diagram follows the transcript evaluation path. It describes an educational prototype, not validated clinical decision-making.
01 / Supply practice material
02 / Organize the presentation
03 / Evaluate
04 / Return and preserve
Control flow / Decisions & data
From practice material to structured feedback
Upload and transcription prepare the input; the evaluation route then transforms a stored transcript into an inspectable educational result.
Scroll across the diagram to follow each branch
Reading the flow
- Audio preparation is separate from evaluation. The evaluation route begins with transcript text loaded from S3.
- Extraction organizes the presentation so subsequent feedback can refer to sections rather than one undifferentiated text block.
- The reasoning tree and rubric output are intermediate and final model artifacts. Their structure supports inspection; it does not prove that the reasoning or score is valid.
- Completed results are persisted to MongoDB and returned to the application. My contribution was the rubrics and AI evaluation logic within the wider team system.
The starting point
Why this project?
Students practicing History & Physical presentations need feedback that connects to the structure and reasoning in their presentation.
My contribution
The work I brought to it.
Worked with the Rutgers Health Hack team on the rubrics and logic used by the AI. The interface, storage, transcription, and wider application are team work.
How it works
From input to output.
The backend supports uploaded material and transcription. The evaluator route reads transcript text from S3, invokes a staged evaluation pipeline, and saves structured results to MongoDB. The evaluator extracts sections, builds a reasoning tree, and generates scoring and feedback.
- Practice transcript
- Extract sections
- Reason and score
- Structured feedback
A design decision
Make feedback inspectable.
Separating extraction, reasoning, and scoring produces structured output that can be examined. This is an educational prototype; structured feedback does not establish clinical or grading validity.
The evidence
Look under the surface.
These links point to the reviewed source revision, so the implementation behind this story stays inspectable.
Current boundaries
Useful work. Honest limits.
- Educational hackathon prototype, not a clinically validated system.
- No autonomous grading reliability is claimed. Cloud integrations require configuration.
Source reviewed October 3, 2026.
Let’s build something
worth protecting.
“A thoughtful conversation can be the beginning of something worth building.”Connect on LinkedIn