Work
What I’ve built. More to come.
StudyPDF
An AI study tool built around one AI study companion named Bo. You upload your course (lectures, slides, scanned notes, even a YouTube lecture), then you talk to Bo. It answers from your own material and builds flashcards, quizzes, practice exams, study guides, cheat sheets, and mind maps inside the chat, as you ask. Everything stays scoped to your course and is cited back to the exact page, so you study from your material and not the open internet.
How it works
One expensive read at upload time. A reasoning model reads the whole course once and pulls out the structure: the topics, the concepts inside each topic, and how they connect. That graph is stored at the course level and de-duplicated across lectures, so lecture five knows it is talking about the same idea lecture two introduced.
Everything hangs off that one graph. When you ask Bo for something, a separate Researcher agent searches your course, and every concept or figure that lands in an answer or an artifact comes from those verified results, never invented. Citations point back to the lecture and the page.
And every time you study, the result writes back: what you got right, what you stumbled on. That running record is the mastery layer, a per-student picture of what you know that compounds across the semester. The model is not the moat. The memory is.
Stack
- TypeScript
- Next.js
- PostgreSQL
- Pinecone
- Anthropic API
- OpenAI API
- Cloudflare R2
- Monorepo
- Self-hosted
- Claude Code
The rest of the infrastructure is boring on purpose. Boring tech is what runs alone.
My role
Solo. I built and run all of it. The AI pipeline, the backend, the database, storage, auth, the growth side, the support inbox. Between March and December 2025, Moritz Kindler was co-founder and ran the UI/UX and brand work. What you see in the product today is largely his visual identity. We ended things cleanly when the logistics stopped working, and I’ve been running it alone again since January 2026.