Globally, heart disease has been the leading cause of death for more than two decades. There is a need to develop intelligent architectures to handle a variety of real life clinical scenarios when a 12-lead ECG is not a viable option. We propose a method using wide and deep CNN architectures to classify cardiac abnormalities from 12, 6, 4, 3, and 2 leads ECGs. These five networks were created for the PhysioNet/CinC Challenge 2021, by the Biomedic2ai team. ECG signals were down-sampled to 100Hz and partitioned with 5-second windows using a sliding 4-second overlap. A one-dimensional deep CNN (1D-dCNN) module was used to preserve sequentially related features embedded in the signals. A feature extraction module was added to the 1D-dCNN, creating a ‘wide and deep modular network’. This framework allows the addition or removal of modules to optimize classification models. We achieved test scores of 0.36, 0.30, 0.31, 0.29, and 0.34 (ranked 23rd, 26th, 26th, 27th, and 22nd out of 39 officially ranked teams) for 12,6,4,3, and 2 leads, respectively, on the hidden test set provided by the challenge. Our model demonstrates potential with the wide modular network. The framework also provides the flexibility to integrate clinical knowledge in the future modules to improve the overall classification performance.