Worked on enhancing deployment workflows and CI/CD reliability for open-source machine learning infrastructure. In the ggml-org/llama.cpp repository, implemented a graceful shutdown mechanism for Docker containers by refactoring the entrypoint script with bash and YAML, ensuring proper signal handling and resource cleanup to prevent zombie processes during runtime. In the huggingface/text-generation-inference repository, updated the nix_build.yaml workflow to enforce semantic versioning for CI Docker images, introducing consistent tag transformations for improved traceability and safer deployments. Focused on DevOps practices, Docker, and GitHub Actions, the work emphasized maintainability, reproducibility, and operational safety across containerized environments without introducing new bugs.