SatBrain
AI study assistant that converts documents into interactive learning materials — summaries, quizzes, flashcards, and visual charts.
Context & Overview
Students and professionals spend significant effort digesting large volumes of material before actual studying begins. SatBrain applies generative AI to that overhead: static documents (PDFs, audio) are automatically converted into active study aids — quizzes, flashcards, and visual charts.
The Challenge
Problem Statement
Learners spend 60% of their study time organizing and summarizing notes rather than studying, which reduces effective retention per hour invested.
Why It Matters
Compressing time-to-study shifts user attention from preparation to comprehension and recall, directly improving learning outcomes.
Constraints
- Zero infrastructure budget
- High-accuracy summaries with low hallucination tolerance
- Near-real-time processing latency
Goals & Metrics
Objectives
- Automate study material creation from raw files.
- Generate visual insights (charts) from document content.
- Keep AI interactions under 5 seconds end-to-end.
Target KPIs
- User retention rate
- Documents processed per user
- Quiz completion rates
Research & Discovery
Methodology
Key Insights
- Students valued visual outputs (charts) nearly as much as text summaries.
- Flashcards were the most requested retention feature.
- Audio transcription is the key differentiator for lecture recordings.
Approach & Strategy
Vertical integration of AI services, built around the Gemini 1.5 family — its large context window allows entire textbooks to be processed in a single pass instead of fragmented chunk-by-chunk inference.
Frameworks
Gemini API for summaries, quizzes, and flashcards (@google/generative-ai), Supabase for auth, storage, and persistence, Radix Themes + Radix primitives for a component-driven frontend
Collaboration
Direct feedback loop with beta testers driving rapid feature iteration.
The Solution
A web platform where users upload source material and receive a tailored study dashboard — document processor, chart generator, and an interactive quiz engine.
App Visuals
Design Rationale
A Vite + React SPA with Supabase and the Gemini API enabled rapid prototyping while providing enterprise-grade auth and database behavior out of the box.
Execution
Roadmap
Phase 1
Document upload & AI summaries
Phase 2
Quiz & flashcard engines
Phase 3
Visual charts & audio input
Challenges Overcome
- ⚠ Large PDF ingestion required a chunking strategy.
- ⚠ Consistent JSON output from the LLM for chart generation was enforced via robust validation middleware.
Outcomes & Impact
Quantifiable Results
- 1.Processed 500+ documents in beta
- 2.Reduced study prep time by ~70%
- 3.90% positive feedback on the visual charts feature
Qualitative Feedback
“Users reported feeling more prepared for exams.”
“The clean, distraction-free UI drew consistent praise.”
Learnings & Future
Key Takeaways
- Generative AI requires strict guardrails when the output is structured data.
- Visualizations drive stronger engagement than plain text summaries.
Next Steps
- Mobile app development
- Collaborative study groups
- Integration with Canvas/LMS