StudyPilot

Inspiration

Students often know their overall marks but struggle to identify which topics they should focus on and what to study next. This inspired us to build StudyPilot, an AI-powered study-planning app that turns topic-level scores into a personalized, reasoning-based study plan.

Instead of simply showing students their weaknesses, StudyPilot aims to answer the more useful question: “What should I study next?”

How We Built It

We designed StudyPilot as a lightweight hackathon MVP, focusing on the core flow rather than unnecessary infrastructure.

The app takes a student's topic-level scores as input and uses an LLM through the Hugging Face Inference API to generate a reasoning-based study plan.

Our planned technology stack includes:

Python + FastAPI + Uvicorn for the backend MongoDB + PyMongo for storing student data, attempts, questions, and generated plans Hugging Face Inference API for AI-powered plan generation HTML5, CSS3, and JavaScript for the frontend Railway for deployment if time allows We followed a simple build philosophy: test the riskiest AI component first, then build the interface around a working response. What We Learned

Through this project, we learned that building an AI application is not only about integrating an LLM. The most important part is designing a useful flow around the AI.

We learned to:

Turn raw performance data into actionable insights. Keep an MVP focused on the feature that provides the most value. Test risky integrations before investing time in the UI. Choose familiar technologies to reduce development time during a hackathon. Distinguish between features that improve the core product and features that only add infrastructure complexity. Challenges We Faced

One of our biggest challenges was the LLM integration, since the AI-generated study plan is the core intelligent component and also the riskiest dependency in our stack. We needed to make sure the model could consistently transform topic scores into useful study-plan text before building the rest of the experience around it.

We also had to make difficult scope decisions. Features such as real authentication, notifications, vision models, maps, embeddings, and social login were deliberately excluded because they did not directly contribute to the core study-planning experience.

Our Goal

With StudyPilot, our goal is simple: help students spend less time figuring out what to study and more time actually studying.

By combining performance data with AI-generated guidance, StudyPilot turns scores into a clear and personalized path forward.