Aspects extend to methods, systems, and computer program products for transforming data manipulation code into data workflow. Data manipulation code for a data science process is written in a data manipulation programming language. The data manipulation code defines input instructions, data manipulation instructions, and output instructions. A learning module automatically transforms the data manipulation code into a data workflow representative of the data science process. The level of detail for a data workflow can be tailored for an intended audience and/or for subsequent editing with an editor program (e.g., a drawing program). Aspects of the invention address the disconnection between designing a data science process and documenting the data science process. The creation of data workflows is automated, virtually eliminating manual operations and providing significant productive gains for data scientists, data engineers, developers, and program managers.