Miguel worked on the autoppia/autoppia_iwa and autoppia/autoppia_webs_demo repositories, focusing on stabilizing local LLM services and enabling scalable, multi-instance web demo deployments. He implemented parallel LLM execution with concurrency control, expanded stress testing frameworks, and improved error handling to ensure reliability under load. Using Python, Docker, and Bash, Miguel addressed dependency management issues, upgraded core libraries, and enhanced observability through improved logging and diagnostics. His work included port-configurable deployments and robust configuration management, allowing multiple demo instances to run concurrently. Across both repositories, Miguel delivered well-tested, maintainable backend systems that improved performance, resilience, and deployment flexibility.