Since you mentioned using a sort, chaining together two Sort components in the Data Flow would work. This would only make sense if SSIS was running on a server different from the source database though, otherwise SSIS would just be competing for resources with SQL.
[OLEDB Source] --> [Sort Component 1] --> [Sort Component 2] --> [OLEDB Destination]
Sort Component 1: Sort by EmployeeID (ASC) and LevelID (DESC).
Sort Component 2: Sort by EmployeeID (ASC) only, passing through Level ID, and select "Remove rows with duplicate sort values" in the Sort Transformation Editor dialog.
I don't see any statement in BOL about which duplicate is retained, but it appears the last one read is copied to the output.