CircuitMentor

About CircuitMentor

The problem

Learning electronics from static diagrams and long text tutorials is slow and easy to misread. It’s hard to picture how parts sit on a bench, which connections cross, or what order to follow.

What we built

CircuitMentor is an interactive tutor: you describe a project, and the system turns that into a step-by-step, narrated 3D walkthrough—a spatial mental model, not just a schematic. An AI presenter can explain each step so it feels closer to working with someone at a bench than reading a PDF.

How it fits together

A user prompt becomes structured wiring and lesson steps (Gemini). Real 3D component models ground the scene in recognizable hardware. Voice (ElevenLabs) and an AI avatar add presentation on top of the instructions. The app is Next.js with React Three Fiber for the scene; FastAPI orchestrates the backend, with MongoDB for persistence.

What was hard

Keeping LLM output, 3D layout, and pin naming aligned so what you see matches what you’re told to wire—and keeping latency acceptable while several external APIs run in the loop.

Takeaways

Multimodal products are not “LLM plus a widget”—they are about glue: consistent schemas, UI that survives messy data, and a client that stays stable when layout and state change. This project pushed us past text-only AI into something you can see, hear, and step through.

Github repo link: https://github.com/PYthegr8/HackDartmouth_TeamMergeConflict/tree/main

Built With

• elevenlabs
• fastapi
• gemini
• liveavatar
• mongodb
• next
• python
• react
• typescript