How I Built the Engine That Makes Our AI Mentor Actually Work

By: Devika N D - Code Execution & Behavioral Signal Module Hindsight Hackathon - Team 1/0 coders *“The moment I submitted an infinite loop and watched the server hang forever — * I realized the execution engine is the most dangerous part of the whole system.” Nobody told me that when you let users run arbitrary Python code on your server, you are one while True: pass away from a dead process. I found out the hard way. My job in this project was building the code execution engine, problem store, behavioral signal tracker, and cognitive pattern analyzer. \This is the story of what I built, what broke, and what I'd do differently. What We Built The AI Coding Mentor is a system where students submit Python solutions to coding problems, get evaluated against real test cases, and receive personalized hints based on how they actually behave — not what they tell us about themselves. The stack: __FastAPI __backend for all routing and execution __Groq (LLaMA 3.3 70B) __for AI-generated feedback