AutoMentor AI

• 

  AutoMentor AI

Computer science and engineering students frequently get trapped in "tutorial hell" or succumb to "AI vending machine syndrome", copying ready-made code without building true mental models. Moreover, studying complex college syllabi (Distributed Systems, Concurrency, Cloud Architecture) lacks hands-on practice, automated feedback, and portfolio visibility. We envisioned AutoMentor AI: a truly autonomous collaborative partner that acts like an empathetic, senior Tech Lead. Instead of writing the code for you, it diagnoses knowledge gaps, schedules study blocks on Google Calendar, scaffolds real GitHub lab repositories, reviews Pull Requests with Socratic precision, and turns student mastery into verified LinkedIn portfolio showcases.

Balancing true Socratic guidance with student momentum was challenging... AI models often want to dump the answer immediately. We engineered production-grade system prompts with strict cognitive guardrails and built the Progressive Hint Ladder. Additionally, orchestrating multi-tool agentic chaining across Google Calendar, GitHub REST APIs, and Firestore required resilient error boundaries and local fallback execution for zero-downtime reliability.

Track Selected: The Collaborative Partner

Built With

• and-@xyflow/react.-client-side-execution:-webassembly-powered-pyodide-runner-for-instant
• and-ebbinghaus-decay-schedules.-frontend-cockpit:-next.js-15-(app-router)
• bloom-taxonomy-levels
• fastapi
• gemini-3.5-flash
• google-adk
• google-cloud-firestore
• google-cloud-run
• google-genai-sdk
• lucide-icons
• monaco-editor
• nextjs-15
• pyodide
• python
• react-19
• tailwind-css
• tool-failures
• typescript
• zero-latency-in-browser-test-validation.-testing:-exhaustive-38-scenario-automated-test-battery-(pytest)-covering-student-cognitive-edge-cases