Walnut

Walnut is an AI-native learning platform built around a simple bet: a tutor that knows your exact course is far more useful than a general-purpose chatbot. Instead of dropping students into an open-ended model, Walnut grounds every answer in the material the professor actually assigned — and gives instructors tools that make grading and feedback take minutes instead of evenings.

What I built

• A student dashboard with assignments, automated grading, and a real-time discussion space where questions and answers stay tied to the course.
• A teacher dashboard that surfaces where a class is struggling, so instructors can spend their time on the parts that need a human.
• A “Contextual Intelligence” engine layered on the OpenAI API. It retrieves from course-specific material and enforces guardrails, so the AI tutor explains and nudges rather than just handing over answers.
• An automated professor toolkit that ingests uploaded readings, slides, and assignments and generates grading rubrics from them.

Why it matters

Most “AI in education” tools are a chat box with a school logo on it. The hard part isn’t calling a model — it’s keeping it accurate, on-syllabus, and academically honest. Walnut is my attempt at that: retrieval and guardrails first, polish second. Building it end to end meant owning the data model, the API design, the React front end, and the prompt and retrieval layer that ties them together.

The pitch

I took Walnut to Oraseya Capital in Dubai and pitched it as a product, not a demo — the problem, the wedge, and the path to a real classroom. It landed: a working demo turned into a real seed conversation. Pitching forced me to defend every assumption out loud, which made the product sharper than any code review could have.