AutoFixee

Architected an enterprise-grade Multi-Agent system using Python and FastAPI, orchestrating Claude 3.5 Sonnet to autonomously diagnose and patch CI/CD failures across heterogeneous codebases. Integrated Greptile API for advanced RAG, utilizing vector-graph indexing to provide deep codebase observability and eliminate file-path hallucinations during the diagnostic and planning phases. Implemented secure execution via E2B cloud sandboxes, automating the validation of AI-generated fixes through unit testing and Docker-based containerization before GitHub PR submission. Developed a self-healing PostgreSQL backend to manage stateful fix histories and persistent agent memory, ensuring high reliability, scalability, and performance for complex technical challenges.