sparkle

AI-powered concept diagnosis and personalized learning paths for students.

Comment

Sparkle

Sparkle is an AI-powered learning diagnosis platform that helps students identify why they are struggling, not just what they got wrong. Instead of treating quiz scores as the final outcome, Sparkle analyzes concept-level understanding across DSA, Python, SQL, and Machine Learning, builds a personalized skill graph, and generates targeted practice and AI-guided explanations for weak areas.

Inspiration

I built Sparkle after realizing that most learning platforms stop at marks and percentages. A student may score 60% in Python, but that score does not reveal whether the problem is functions, modules, recursion, or problem solving. I wanted to create something that behaves more like a mentor—one that diagnoses conceptual gaps and recommends the shortest path to mastery.

The idea came from my own experience preparing for technical interviews and university exams, where I often spent time practicing questions without knowing which concepts were actually holding me back.

What it does

Sparkle combines adaptive quizzes, concept mapping, and an AI mentor into a single workflow:

  • Randomized quizzes for multiple technical subjects
  • Concept-level mastery tracking instead of only overall scores
  • Weak-concept detection across all subjects
  • A skill graph that visualizes relationships between concepts
  • AI-generated explanations and personalized next-step recommendations
  • Dashboard analytics showing mastery trends and learning progress

The goal is to help students focus on the concepts that will produce the highest improvement in the least amount of time.

How I built it

Frontend

  • Next.js
  • React
  • TypeScript

I built a modular dashboard, quiz interface, skill graph view, and AI mentor interface with reusable components and responsive layouts.

Backend

  • FastAPI
  • SQLAlchemy
  • Alembic
  • SQLite

The backend exposes APIs for authentication, quizzes, dashboards, diagnosis, and AI interactions.

AI and Recommendation Layer

I designed a concept graph representing prerequisite relationships between topics and built a diagnosis engine that maps quiz performance to concept mastery. Sparkle uses AI to generate personalized explanations and remediation strategies based on a student's weakest concepts.

The backend also seeds a structured knowledge graph and question bank, allowing the system to generate subject-specific insights rather than simple score summaries.

Challenges I ran into

One of the biggest challenges was designing a diagnosis engine that could infer conceptual weaknesses from quiz attempts instead of relying only on final scores. I also had to keep the frontend and backend architectures clean while rapidly iterating during the hackathon.

Other challenges included:

  • Structuring the concept graph and prerequisite relationships
  • Handling randomized quiz generation consistently
  • Designing a dashboard that remains understandable while displaying concept-level analytics
  • Integrating AI recommendations without making the experience feel slow or overwhelming

What I learned

Building Sparkle solo taught me how to combine AI, knowledge graphs, and educational analytics into a practical product. I gained hands-on experience with:

  • FastAPI application architecture
  • Database migrations and schema design
  • Next.js dashboard development
  • AI service integration
  • Graph-based recommendation systems
  • Building a full-stack product independently under hackathon time constraints

The biggest lesson was that meaningful educational AI is not only about generating answers—it is about generating useful diagnoses and actionable learning paths.

What's next for Sparkle

I plan to expand Sparkle into a broader adaptive learning platform by adding:

  • More subjects and deeper concept graphs
  • Spaced repetition scheduling
  • Interview-mode assessments
  • Collaborative learning features
  • Analytics for educators and mentors
  • Cloud deployment with persistent user data

Sparkle is my attempt to make technical learning more personalized, measurable, and efficient—helping students spend less time guessing what to study and more time mastering the concepts that matter.

Share this project:

Updates

Submission history