@nkrish: Thank you for the response

If you don't need Spark parallelism, can you review these options?

Option-1: APIfy the exe in a separate windows vm/container
       - call the API from Databricks. this call will be executed from the driver. 

Option-2: Recompile exe for Linux        
        - compile C# for Linux platform
        - install .NET runtime for Linux in Databricks Runtime using init script
        - copy the exe from storage to cluster using init script
        - execute the exe using %sh or python subprocess or similar approaches in the notebook

Both are anti-patterns. Option-1 is less invasive and will take less efforts compared to Option-2.

Once you review and understand internals of exe, think about rewrite/porting to Spark APIs for better maintenance and performance.