Inspiration
As developers and learners, we noticed a gap between real-world coding experience and online learning resources. Many platforms teach concepts in isolation, without connecting them to actual projects or measurable progress. We wanted to bridge this gap by creating an AI-powered platform that analyzes real GitHub repositories and provides actionable learning insights.
What it does
MANTIS helps developers and learners by:
Analyzing GitHub repositories for code quality, documentation, and maintainability
Generating personalized, job-ready learning roadmaps
Providing AI-driven code evaluation and mentoring
Tracking skill progression and learning analytics over time
Producing AI-generated developer profiles and resume signals
How we built it
Backend: Python, Flask, SQLite
AI Layer: ERNIE 4.5 models via Novita AI
Features:
GitHub repo analysis → structured JSON scores
Learning roadmap generation → progressive skill-based items
AI code evaluation → feedback, improved code, and scoring
Data Flow: Repo download → file parsing → ERNIE analysis → structured feedback
We focused on structured AI reasoning instead of just generating text, ensuring accurate, evidence-based feedback.
Challenges we ran into
Handling large repositories and ensuring models could reason over multiple files
Designing strict JSON contracts to guarantee reliable AI outputs
Integrating multiple ERNIE models for different tasks while keeping the system efficient
Providing meaningful, actionable learning insights rather than generic feedback
Accomplishments that we're proud of
Successfully built a platform that analyzes real GitHub repositories end-to-end
Implemented multi-model ERNIE reasoning for feature-specific tasks
Created job-ready learning roadmaps based on actual repository content
Delivered AI-powered code evaluation with structured scoring
Turned GitHub activity into measurable learning intelligence
What we learned
How to use multimodal AI models (ERNIE 4.5) effectively in real applications
Importance of structured prompting and JSON contracts to avoid AI hallucinations
Challenges of scaling AI analysis for full repositories versus individual code snippets
How to translate AI insights into actionable learning and skill progression
What's next for MANTIS
Add auto PR review assistance for real-time code improvement
Implement team dashboards for classrooms and collaborative learning
Enable multi-agent curriculum planning for advanced personalized roadmaps
Generate resume-ready PDFs from AI-analyzed repositories
Deploy cloud version with CI/CD integration for seamless updates
Log in or sign up for Devpost to join the conversation.