Worked on the linkedin/ambry repository to deliver targeted backend improvements focused on reliability and observability. Developed resilient named blob retrieval by introducing configurable error translation, allowing BlobDoesNotExist errors to be mapped to retryable responses based on per-cluster settings. Enhanced error handling and reduced client retry storms by returning appropriate HTTP status codes when storage replicas were unavailable. Centralized JSON handling through a new utility to ensure consistent ObjectMapper configuration, supporting safer rolling deployments. Additionally, improved log message formatting to align with semantic versioning, aiding debugging and incident response. Utilized Java, API development, and version control to implement these changes.