Remi | Devpost

Inspiration

In today’s hyperconnected world, staying focused on long-term goals has become harder than ever. Social media platforms like Instagram, YouTube, and TikTok have led to skyrocketing screen times, leaving less room for meaningful learning. For example, U.S. TikTok users’ monthly screen time has grown dramatically from 2020 to 2024, reaching staggering new heights.

Our project, Remi, was born out of a desire to reclaim that time. We wanted to turn moments of distraction into moments of productivity — transforming everyday waiting time into learning opportunities.

According to recent ACM findings (source):

Waiting time usage: 60% leisure, 20% productive, 20% maintenance.

This distribution shows tremendous potential: by reimagining how we use fragmented moments, we can greatly boost productive and educational time. That’s what Remi does — it converts your “dead time” into “learning time.”

What It Does

Remi is an iMessage-based learning assistant that helps users actively engage with their study materials — seamlessly integrated into their daily lives.

Upload and Automate: Users upload course materials (PDFs, notes, or text). Using OAuth, Remi connects to their Google Calendar to find available free time slots.
Smart Scheduling: It automatically creates a study plan, evenly distributing review sessions based on topic intensity and preferred frequency.
Interactive Recall: During these sessions, Remi sends review questions through iMessage — no need to open another app or stop what you’re doing.
Performance Feedback: The platform generates progress reports and weekly/monthly summaries.

Features

Dynamic learning feedback based on question accuracy.
Analytics dashboard displaying correct/incorrect answers, topic progress, and learning curves.
Adaptive scheduling that adjusts as your knowledge and availability evolve.
Seamless iMessage integration — no new app downloads needed.

How We Built It

Frontend: Built with React.js for a responsive and modern web experience.
Backend: Powered by Express.js, managing communication between the frontend, databases, and the Photon iMessage server.
Messaging Engine: Implemented via Photon’s open-source iMessage Kit, enabling real-time, bidirectional message exchange.
AI Model: Gemini serves as the LLM, chosen for its large context window (up to 1M tokens) — perfect for understanding lengthy course PDFs.

Challenges We Overcame

One of our biggest challenges was integrating our frontend, backend, and iMessage system into a single cohesive product. We had three distinct components: a React frontend, an Express.js backend, and an iMessage integration powered by Photon. Bringing these systems together was complex, but we overcame it by dividing tasks strategically. One team member focused on the frontend, another handled the backend, and a third worked on connecting Gemini and Photon. This collaboration allowed us to work efficiently and ultimately build a functional, user-ready product.

Accomplishments We’re Proud Of

Built a fully functional multi-agent iMessage learning platform.
Overcame complex distributed system challenges.
Developed an open-source foundation for iMessage-based active learning.
Designed a clean and data-driven frontend for progress visualization.

What We Learned

How to integrate three servers (frontend, backend, and Photon iMessage) into a distributed architecture.
The fundamentals of Express.js and NoSQL database management.
Techniques for managing stateful multi-session AI conversations.
The importance of contextual design in educational technology.

What’s Next for Remi

Canvas Integration: Automate course material uploads for college students.
Calendar Integration: Factors in schedule to decide when questions are sent.
Creative Learning Expansion: Support for musicians (upload sheet music, receive AI feedback on recordings).
Visual Learning: Enable photographers to upload images and receive feedback on technique.