LearnBridge

Inspiration

We’ve all experienced the stress of being late to class and scrambling to find the right course, unit, or worksheet, only to be slowed down further by long loading times and clunky interfaces. This repeated frustration not only wastes time but also creates anxiety and distracts from the learning process. We wanted a better, faster way to access learning content without the panic.

What it does

LearnBridge is an AI-powered chatbot built with Google ADK that helps students instantly locate assignments, posts, and course materials from platforms like Google Classroom and D2L. By streamlining access to content, it reduces frustration, saves time, and improves productivity for both students and teachers.

How we built it

Language & Runtime: Python 3.12

APIs & Auth: google-auth-oauthlib and google-api-python-client for Google Classroom integration. As well as the Google Calender API.

Data Handling: requests, pytz, and python-dateutil for API calls and timestamp normalization

User Interface: A clean, user-friendly interface (non-command-line) with Streamlit that simplifies interaction with the chatbot

We used Google ADK with the Google Deepmind Gemini LLM Model to build a multi-agent system that allows the bot to carry out intelligent, context-aware interactions with users.

Challenges we ran into

One of the main challenges was dealing with Google Classroom's OAuth restrictions, particularly redirect URI rules and token refresh handling. Setting up a reliable login flow took time, especially while managing edge cases like expired or invalid tokens.

Accomplishments that we're proud of

Learning and applying Google ADK from scratch

Designing a multi-agent system that mimics real-world assistant interactions

Building a tool that we not only believe in, but actually use in our daily academic life

What we learned

Every LMS comes with its own authentication quirks, redirect URIs, scope limitations, and token refresh rules all behave differently. We quickly realized the value of designing a robust, user-friendly auth flow from day one to save hours of late-night debugging.

Beyond integration, we gained hands-on experience with Google ADK and learned how to build smart, modular workflows using agent state memory, as well as both parallel and sequential task execution. This allowed us to design fluid conversations and intelligent task management.

We also explored how different LLMs behave under the same prompts, and how prompt engineering and model selection can drastically change performance. We were genuinely surprised by how powerful LLMs can be when given clear, well-structured instructions, they really are capable of much more than expected when used intentionally.

What's next for LearnBridge

We aim to integrate LearnBridge directly into other Learning platforms like D2L and Duolingo to crreate a central learning spot that will greatly improve usability and enhance the daily experience for both students and educators, no matter their current location, knowledge, and setup.

Our agentic architecture was designed with scalability in mind, making it easy to expand and support new use cases. In the future, we plan to integrate features like Google Meet access, reminders and to-do list management, and even campus maps for planning when to leave for class. Our goal is to become an all-in-one AI companion for students, helping them stay organized, on time, and focused on what matters most. It’s about boosting productivity, efficiency, and performance across the board.

Built With

• adk
• and-even-campus-maps-for-planning-when-to-leave-for-class.-our-goal-is-to-become-an-all-in-one-ai-companion-for-students?helping-them-stay-organized
• and-focused-on-what-matters-most.-it?s-about-boosting-productivity
• calender
• classroom
• efficiency
• embedding-our-chatbot-into-the-learning-environment-will-greatly-improve-usability-and-enhance-the-daily-experience-for-both-students-and-educators.-our-agentic-architecture-was-designed-with-scalability-in-mind
• gemini
• google
• making-it-easy-to-expand-and-support-new-use-cases.-in-the-future
• on-time
• python
• reminders-and-to-do-list-management
• streamlit
• we-plan-to-integrate-features-like-google-meet-access