Back to selected work

Full-stack ML application

Student Performance Predictor Pro

An authenticated student-performance application connecting prediction, uncertainty, and recommendations through a focused web interface.

What it is

An authenticated student-performance application connecting prediction, uncertainty, and recommendations through a focused web interface.

What it demonstrates

A FastAPI backend exposes health, prediction, uncertainty, recommendation, and authenticated-user endpoints for a React and TypeScript client.

My role

Independent builder

Project status

Public application demo

01The problem

A prediction interface needs more than a score: people need an understandable workflow around uncertainty, follow-up, and the appropriate use of recommendations.

02System architecture

SYSTEM MAP

React and TypeScript interface
FastAPI backend
Prediction endpoint
Uncertainty endpoint
Recommendation endpoint
Authenticated user flow

03Model and product pipeline

FLOW

User input
Validation
Prediction
Uncertainty signal
Recommendation
FastAPI response

04Engineering decisions

01

Expose uncertainty separately

The API provides a dedicated uncertainty path instead of implying an unqualified answer.

02

Keep recommendations downstream

Recommendations are a separate workflow from the prediction response.

03

Use authenticated flows

Sign-up, login, and current-user endpoints scope the product experience.

04

Keep the client focused

React, TypeScript, and Vite support a clear interface around the API.

05Product screenshots

Student Performance Predictor Pro public application landing interface
Live public application: prediction, uncertainty, and intervention-planning entry point.
Student Performance Predictor Pro assessment flow
Live public application: a multi-step assessment view with prediction-ready signals.

06Results and evaluation

  • - The public repository documents the implemented API surface and deployment path.
  • - No accuracy or educational-outcome metric is claimed on this portfolio page.

07Limitations

  • - This is not a validated educational intervention.
  • - Predictions and recommendations require appropriate human review and context.

08What I learned

Uncertainty deserves a first-class product surface.

An API contract can make a model workflow easier to test and evolve.

A useful UI states what a system cannot conclude.

PythonFastAPIReactTypeScriptVite