Worked on the ai-dynamo/dynamo repository to refactor the TRT-LLM deployment architecture, focusing on optimizing model serving and resource utilization. The approach involved changing the component type from 'main' to 'worker' across multiple deployment configurations, enabling more efficient use of worker nodes for TRT-LLM processing. This adjustment laid the foundation for improved scalability and better alignment with cloud deployment best practices. The work was implemented using Kubernetes and YAML, leveraging DevOps principles to ensure traceability through a single commit and pull request. No bugs were addressed during this period, with efforts concentrated on delivering this targeted architectural enhancement.