Network editing and model experimentation are critical processes in the development of deep learning (DL) applications. This paper presents an IDE that supports these two processes for novice DL programmers. It consists of a network editing mode and an experimenting mode, which correspond to the two processes of DL application development. The network editing mode utilizes the live programming technique to synchronously provide tensor shape visualization when the user modifies the network structure. The experimenting mode then takes the edited network structure and helps manage the DL model experimentation. Novices can conveniently transit between network editing and experimentation to iteratively improve model performance with the tool.