CUIDE
A SystemVerilog learning environment for university students
Motivated by the confusion of many Cornell professors over what to do with AI, I created CU IDE!
Students write RTL in a Monaco editor, run visible tests, and submit their work to a separate hidden test harness. A chat-based AI teaching assistant can inspect their current code and offer conceptual guidance, with guardrails designed to prevent it from writing or completing the solution.
technical work
- Built a React and TypeScript interface with a Monaco editor and IBM Carbon components.
- Implemented a Rust API and an Icarus Verilog grader running in Docker with network access disabled and resource limits.
- Added a teaching assistant with bounded context, response validation, and checks against protected reference material.
- Built instructor views for reviewing code snapshots, student interactions, and TA conversations, with on-demand student and cohort analysis backed by SQLite.
This is a local prototype; classroom pilot testing and production deployment are still ahead.