All projects

LearnMate AI

A self-learning companion that generates a learning path for any skill, measures focus and rewards consistency.

Kind
Full-stack web app
Year
2025
Recognition
1st in Karnataka · Rewind & Recode, IIIT Bhubaneswar, 24-hour build
Built with
  • Next.js
  • React
  • TypeScript
  • Gemini
  • Firebase
  • Tailwind CSS

The question

What would a study plan look like if it was built around a learner’s own goal, and kept them coming back?

LearnMate turns any skill into a path, then supplies the structure and the nudges a teacher would.

Why it exists

Self-directed learning has no structure and no one checking in. LearnMate builds the path, tracks engagement such as watch time and streaks, and adds points, challenges and study buddies.

How it works

From a skill to a streak

  1. 1

    Goal

    You name a skill and pick a style, such as theory-heavy or project-based.

  2. 2

    Path

    Gemini returns five to eight steps, each with two to four resources.

  3. 3

    Study

    Videos play in an embedded player that tracks watch time and focus, and you can take timestamped notes.

  4. 4

    Reward

    Completing resources and steps earns points and extends your streak, and the leaderboard and challenges respond.

Decisions

  • Streaks are worked out from the sessions themselves

    The current streak is counted from completed sessions, grouped by calendar day, each time the summary is read, so it follows the study history directly.

  • Quizzes are graded in code

    Gemini writes the questions, but submissions are graded by plain logic, so a score does not depend on the model.

  • Fallbacks where a model or an index may be missing

    The leaderboard falls back to all-time standings if its index is missing, and study-buddy matching falls back to a heuristic score if the model’s answer is absent.

Trade-offs and limits

  • Focus is measured from tab visibility, which shows whether the video stayed in view rather than whether the learner was paying attention.

Details

  • 61 API handlers across 37 route files
  • 23 composite Firestore indexes

Built with

Languages

  • TypeScript

Frameworks and libraries

  • Next.js15
  • React19
  • Tailwind CSS4

Data and storage

  • Firebase

AI and models

  • Gemini