StudySketch AI — turn notes into diagrams & mind-maps 📚🖊️

Inspiration 💡

Students and professionals often deal with dense notes, long PDFs, and unstructured lecture materials. We wanted to make studying faster, visual, and interactive by converting these materials into diagrams, mind-maps, and flashcards — all on-device for privacy and speed.

What it does 🛠️

Upload PDFs, DOCX, TXT, EPUB, PNG/JPG (OCR supported). 📄🖼️
Smart parsing of headings, bullets, tables, and semantic sections. 🔍
Auto-generation of hierarchical mind-maps from text. 🌳
Diagram creation: flowcharts, concept maps, causal graphs. 🗺️
Multi-level summaries: one-liners, paragraphs, or page-level. ✏️
Interactive editor: edit nodes, merge/split branches, customize colors/layout. 🎨
Flashcards & spaced-repetition Q/A export (CSV/Anki). 🃏
Export as PNG, SVG, OPML, JSON, Markdown, PDF. 💾
Fully offline and private: all AI runs locally. 🔒
Performance mode: switch between high-quality or low-latency inference. ⚡
Semantic search over uploaded notes and generated diagrams. 🔎

How we built it 🏗️

Frontend: Flutter (iOS + Android), React Native optional.
iOS Native: Swift + Core ML for Apple Silicon optimization. 🍏
Model Pipeline:
- OCR via Tesseract / OCR-Mobile for images and PDFs.
- Text cleanup, semantic chunking, and summarization via small quantized transformer.
- Graph generation: transformer or rule-based node/edge JSON output.
Storage: Local SQLite + light-weight on-device embeddings (FAISS).
Export: SVG/PNG renderer, OPML, JSON.
Apple/Arm optimization: Core ML quantized models, NEON acceleration, on-device GPU/NPU via Core ML delegates. ⚙️

Challenges we ran into ⚠️

Ensuring on-device AI ran efficiently on Apple devices without draining battery. 🔋
Optimizing OCR and text parsing to handle diverse document formats. 📑
Maintaining smooth interactive mind-map editing while processing large documents. 🖌️
Converting transformer models to Core ML and quantizing without losing accuracy. 🤖

Accomplishments that we're proud of 🏆

Fully functional on-device workflow from document → mind-map → flashcards.
Interactive, editable diagrams with export options across multiple formats.
Optimized for Apple Silicon & Arm devices, giving fast inference and low power usage. ⚡
Open-source ready with clear model conversion and build instructions. 💻

What we learned 📖

The importance of quantization and model optimization for mobile AI.
Balancing UI interactivity with AI-heavy background processing.
How on-device AI enables privacy and instant access, without cloud reliance. 🌐❌

What's next for StudySketch AI 🚀

Add real-time collaborative editing for study groups. 👥
Integrate speech-to-text OCR for lecture recordings. 🎤
Extend diagram types: timelines, cause-effect maps, and interactive quizzes. ⏳
Build iPad-specific UI with Apple Pencil support for direct annotation. ✏️🖌️