## The problem

Language learners need regular practice with people who speak the language they are learning, but finding a good partner and agreeing on times takes effort.

## Key decisions

Recommendations that learn from likes

Each user keeps an "ideal match" profile: the running average of everyone they have liked. Candidates are ranked by weighted distance from it, with shared interests weighted most, and people who already liked the user are shown first to maximize matches.

Scheduling without back-and-forth

Booking a session creates a Google Calendar event and invites the partner, so both calendars stay in sync.

Calls inside the app

Video calls run on Agora, with the backend issuing short-lived access tokens for each call.

Tested from the start

Jest tests with mocked Google APIs cover the backend, and GitHub Actions runs them on every change.

## Results

• ✓Best Project Award, CPEN 321
• ✓Recommendations returned in under 150 ms across 10,000 test requests

## Screenshots

## Built with

Java, Android, Node.js, MongoDB, Jest, GitHub Actions, Azure