Jialei Chen enhanced CI/CD reliability and build stability across the Intel-tensorflow/xla and ROCm/tensorflow-upstream repositories by implementing automated free disk space actions within Bazel workflows, addressing persistent “No space left on device” errors. Leveraging C++ and YAML, Jialei refactored include paths and improved dependency management in the StableHLO Builder, introducing stablehlo_broadcast_lowering to expand broadcast operation support. The work aligned headers with LLVM SmallVectorExtras, ensuring consistent pipeline behavior and reducing build friction. By focusing on workflow automation and DevOps best practices, Jialei delivered robust cross-repository improvements that reduced flaky builds and enabled more reliable, broadcast-capable TensorFlow integrations.