Worked on improving the reliability of AWS Batch integration within the pytorch/torchx repository by addressing a specific bug affecting single-replica workloads. Focused on ensuring that the instance type is always included in the AWS Batch job definition, even when only one replica is specified, which previously led to misconfigurations and support issues. Utilized Python and cloud computing skills to implement the fix and expanded the test suite to cover this edge case, validating changes through continuous integration. This targeted update enhanced the stability of batch job submissions for TorchX users and contributed to a reduction in related support tickets.