Abstract: One or more techniques and/or systems are provided for metadata driven loading a non-uniform multi-row text file. A non-uniform multi-row text file may comprise one or more disparate rows, such as a first row with a name field, an address field, and an age field and a second row with an occupation field, a marital status field, and a gender field. The non-uniform multi-row text file may be evaluated by generating task execution code for tasks identified within a profile for a file type of the non-uniform multi-row text file (e.g., dynamically generate structure query language commands may be executed based upon the task execution code). Responsive to processing the task list, parsed output data for the non-uniform multi-row text file may be generated based upon result data for the tasks. The parsed output data may be outputted into a staging area, such as a staging table, for client access.