College Copilot

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

College Copilot is an Android AI assistant that helps college students manage their academic life, working both offline and online with local LLMs.

I built it for my friend Vishal, who has a common but frustrating problem: he often forgets his college timetable, important submission dates, assignments, and other academic tasks. Even when he writes everything down in his notes, he sometimes forgets to check those notes at the right time.

Vishal can simply take a photo of his college timetable, and the app can understand and store the schedule. He can then set important dates and times for classes, submissions, assignments, and other tasks, with alarms and reminders so he doesn't have to constantly remember to check his notes.

He can also simply ask the assistant questions such as:

"What do I have tomorrow?"
"When is my next Maths class?"
"What should I study tonight?"
"When is my assignment submission?"

The goal is simple: instead of expecting a student to remember everything, College Copilot remembers the important things so they can stay on track.

Demo

https://drive.google.com/drive/u/0/folders/1aRicOdmj9hl6C1JKV7l6X-z6KOhbSHih

Code

vinayak29-ux / College-Copilot

How I Built It

I built College Copilot specifically for Android, taking inspiration from the open-source llama.cpp project for running LLM inference locally on-device.

For the local AI, I used the open-weight Gemma 3 4B IT Q4_K_M model. This allows the app to handle normal conversations directly on the device, making the core experience available even without an internet connection.

For the more advanced College Copilot Assistant, I integrated Backboard.io as the agent and memory layer. I used its persistent memory and retrieval capabilities so the assistant can remember useful information about the student—such as study preferences, routines, and academic context—and use that information in future conversations. Backboard also allows the assistant to combine memory and external information when a question requires a broader or more up-to-date answer.

The result is a hybrid AI architecture: local Gemma handles private, everyday conversations offline, while Backboard + a stronger cloud LLM powers the intelligent assistant experience when online.

Why Does Open Innovation Matter?

Open innovation allowed me to build College Copilot without depending entirely on closed AI APIs.

Using open-weight Gemma enables local, private, offline AI, while Backboard provides memory, retrieval, and stronger cloud-based reasoning when needed.

The key idea is simple: use local AI for everyday tasks and cloud AI when deeper context or capabilities are required.

My Agent Session

Prize Categories

• Best Use of Gemma — for running the open-weight Gemma 3 4B model locally on Android for offline AI.
• Best Use of Backboard.io — for persistent memory, retrieval, and the intelligent Assistant layer powering College Copilot.