Harsh Motla focused on backend development for the NVIDIA/garak repository, addressing a multiprocessing safety issue in the GoogleTranslator integration. Using Python and leveraging multiprocessing techniques, Harsh engineered a solution that prevents the Google Cloud translation client from being pickled, instead instantiating a new client within each worker process. This approach ensures reliable parallel translation by eliminating runtime failures associated with client serialization. Harsh validated the fix through targeted and end-to-end tests simulating real-world multi-process workloads, and updated internal documentation to clarify the new client lifecycle. The work demonstrates a thoughtful approach to robust, scalable backend engineering in Python environments.