← Back to the space · AI systems & research
Healthcare AI platform
Predicting appointment no-shows
- Status
- Completed
- When
- Spring 2026 · Engineering Cloud Computing
- Built with
- Next.js · TypeScript · Tailwind CSS · Node.js · Express · PostgreSQL · Python · FastAPI · scikit-learn · Pandas
A three-tier microservice system that predicts which patients are likely to miss appointments, so a clinic can act before the revenue is lost. A dashboard flags high flight-risk patients, backed by a Node.js orchestrator and a Python service that serves a scikit-learn Random Forest model.
What I built
- Presentation: a Next.js 15 dashboard with client-side filtering and risk-tiered alerts for patients whose predicted no-show probability is above 0.7.
- Logic: a Node.js and Express orchestrator that runs the PostgreSQL queries and calls the AI service over REST.
- Intelligence: a FastAPI service wrapping a serialized scikit-learn Random Forest. Inputs include the lead time between booking and appointment, age and gender, and prior no-show history when available.
- Data: a PostgreSQL schema with Admin, Staff and Doctor roles ready for access control, plus a seed script that generates 600+ mock patient records. No real patient data.
Next: TALON Chat →