We present a search for galaxy-scale strong gravitational lenses in the initial 2500 square degrees of the Canada-France Imaging Survey (CFIS). We design a convolutional neural network (CNN) committee that we apply on a selection of 2344002 exquisite-seeing r-band images of color-selected luminous red galaxies (LRGs). Our classification uses a realistic training set where both the lensing galaxies and the lensed sources are taken from real data, i.e., the CFIS r-band images themselves and Hubble Space Telescope (HST). A total of 9460 candidates obtain a score above 0.5 with the CNN committee. After a visual inspection of the candidates, we find a total of 133 lens candidates, among which 104 are completely new. The set of false positives mainly contains ring, spiral and merger galaxies and to a smaller extent galaxies with nearby companions. We classify 32 of the lens candidates as secure lenses and 101 as maybe lenses. For the 32 best-quality lenses, we also fit a singular isothermal ellipsoid mass profile with external shear along with an elliptical Sersic profile for the lens and source light. This automated modeling step provides distributions of properties for both sources and lenses which have Einstein radii in the range 0.5"<{theta}_E_<2.5". Finally, we introduce a new lens/source single-band deblending algorithm based on auto-encoders representation of our candidates. This is the first time an end-to-end lens-finding and modeling pipeline is assembled together, in view of future lens searches in single band, as will be possible with Euclid.