/ Work · EdTech
A privacy-first AI teaching assistant for universities and professors: a trusted extension of a specific professor, not a generic chatbot.

The client
Zoheb, PhD in Chemistry (Northwestern), AI Fellow at Handshake
He saw universities restricting or banning AI tools, even among people who understand the technology.
The problem
Professors do not trust generic AI tools. Institutions worry about student data, intellectual property and vendor lock-in.
Professors do not want an AI tool; they want a trusted extension of themselves.
What we built
Professor-controlled
- Upload notes, PDFs and readings
- Set tone, teaching style and what the AI may discuss
- Brand the companion and share one course link
Privacy as architecture
- No unnecessary logging, no data crossing between courses
- No model training on student data
- Super admin sees counts only, never conversations or content
Stack React · Node.js · PostgreSQL · OpenAI · RAG · Vector database



Outcome
Lockedprivacy-first architecture
3fronts with hard boundaries
Buildingearly builds in progress
This product is still in development; no outcome metrics yet.
Harder than consumer AI: skeptical expert users, high-trust buyers, and no room for a black box.