Worked on backend systems for ai-dynamo, focusing on stability and deployment flexibility across two repositories. In ai-dynamo/nixl, addressed resource management by fixing a double destruction bug in batch I/O, ensuring proper cleanup and resolving context version mismatches to improve runtime reliability. For ai-dynamo/dynamo, developed dynamic HTTP host configuration for the frontend service, enabling seamless deployment across environments by supporting both CLI flags and environment variables. Leveraged C++, CUDA, and Rust bindings to implement robust backend logic, with attention to safe resource handling and flexible configuration. The work demonstrated depth in backend development, bug resolution, and cross-environment deployment practices.