What it does
Students upload PDFs, slides, YouTube videos, or lecture audio. MyStudySidekick turns them into summaries, quizzes, flashcards, and study plans, and schedules review with spaced repetition.
AI pipeline
- Ingests PDFs, slides, YouTube captions, and audio (Whisper transcription).
- Chunking, embeddings, and retrieval feed LLM generation of study material.
- A SymPy/SciPy computation service for step-by-step STEM problem solving.
Agentic backend
The backend is agentic: instead of single prompt-in, text-out calls, LLM agents work through study tasks using tools and skills.
- Verification tools: agents check generated summaries, quizzes, flashcards, and solutions with dedicated verification tools before content reaches students.
- Skills: reusable, task-specific skills that the agents draw on to produce each type of study material.
Backend
- Go (net/http), PostgreSQL via pgx, Goose migrations, Cloudflare R2 object storage.
- APIs and schema designed so new content types reuse existing library sync instead of adding endpoints.
Clients
- Local-first, offline-capable React Native/Expo app with SQLite.
- Syncs across phone, tablet, and a Tauri desktop app.
- FSRS spaced-repetition scheduling.
Ownership
I own deployment, releases, and production debugging.