Kapil Jain developed a CUDA-enabled debug variant for the llm-d repository, focusing on enhancing debugging workflows for containerized machine learning environments. He updated the build workflow, Dockerfile, and supporting scripts using Shell and YAML to enable debug symbols and verbose logging throughout the build process. By refining cache controls and standardizing debug image naming, Kapil ensured reproducible and deterministic debug builds. His work reduced the mean time to root cause for CUDA-related issues and improved developer productivity. The changes were validated end-to-end in CI and development environments, demonstrating depth in CI/CD, DevOps, and Docker-based engineering practices.