Worked on gpu-mode/discord-cluster-manager to deliver Nvtriton support and integrate the TileIR backend within the Helion Dockerfile, focusing on enhancing performance and compatibility for NVIDIA NVTRITON workflows. The approach involved replacing the upstream Triton backend with TileIR, enabling the TileIR compiler by default through Dockerfile configuration. Validation included end-to-end testing on B200 hardware, confirming kernel correctness for both Triton and Helion matmul operations with zero discrepancies. Leveraging skills in backend development, Docker, and machine learning, the work resulted in a robust test suite outcome and established a foundation for broader deployment and improved reliability in production environments.