MotivateAI

An autonomous AI learning companion that builds sustainable habits by dynamically scheduling sessions, sourcing real-world tutorials, and adaptively managing breaks using Gemini 2.0.

Comment

What it does## 💡 Inspiration

Self-directed online learning is fundamentally broken; over 95% of self-learners drop out of online courses. Traditional platforms provide great video libraries but lack accountability, while standard habit trackers track completion without any cognitive or behavioral depth. They don't care if a user is cruising through a concept in two minutes or hitting a wall with a coding bug for hours. We wanted to build a true cognitive learning companion that acts as an always-on human tutor—affordable, highly personalized, and scalable.

🧠 What it does

MotivateAI is an autonomous AI learning coach that analyzes real-time behavioral signals to help self-directed learners build sustainable work habits.

Users simply type in a learning goal (e.g., "Learn Python encapsulation" or "CS50P Lecture 3 Exceptions"). The system reviews their past completion metrics, average focus durations, and drop-off thresholds to generate a custom, micro-session curriculum. It automatically sources highly relevant educational videos via the YouTube API, matches them with specialized AI-generated Coach Notes, manages active countdown task timers, and provides empathic check-ins and structured breaks calibrated to fatigue patterns.

🛠️ How we built it

  • Frontend: Next.js (App Router), React, and TypeScript wrapped in a beautiful, responsive dark glassmorphic UI utilizing Tailwind CSS.
  • Database: MongoDB for robust, scalable storage of session logs, behavioral metrics, and profile optimization history.
  • AI Core: Google Gemini 2.0 Flash (via the @google/generative-ai SDK), leveraging structured JSON schemas (responseSchema) to guarantee perfect programmatic type safety on the client side, alongside clean system instruction partitioning.
  • Analytics: Visualized history logs powered by Recharts, charting weekly consistency trends, peak performance windows, and task difficulty distributions.

🚧 Challenges we faced

Managing the transition of behavior between strict programmatic generation (creating valid, safe lesson JSON schemas) and open-ended emotional coaching (the conversational greetings) required deep prompt isolation. Ensuring that user input could never override the internal core logic meant strictly partitioning our system instructions to block injection attacks. Additionally, tuning the YouTube Data API search queries to accurately pull premium academic lectures rather than shallow clickbait required robust prompt engineering within the agent's sourcing loop.

🎉 Accomplishments that we're proud of

We are incredibly proud of our "How We've Optimized for You" Engine. Instead of just showing a static graph of user performance, users can trigger a Gemini-driven behavioral audit over their last 30 sessions. The agent writes clear clinical reasoning detailing the user's focus patterns and automatically applies updated optimal break and task durations directly to their live database preferences.

📚 What we learned

Building MotivateAI taught us how crucial strict schema validation is when using LLMs for core UI workflows. Relying on Gemini 2.0 Flash's robust schema enforcement allowed us to confidently drive visual elements like interactive timers, customized progress meters, and dynamic menus without the risk of parsing crashes.

🔮 What's next for MotivateAI

We plan to introduce collaborative "Study Guilds," where multi-agent frameworks match compatible learning profiles together for peer accountability. We also aim to expand signal tracking to monitor browser tab switching or IDE focus to refine our fatigue-detection algorithms even further.

Built With

Share this project:

Updates