A fully automated single cell surgery system is proposed in this thesis. The proposed system firstuses convolutional neural network to detect the embryo location under the microscope field of view. The network produces a bounding box output which is used as region of interest for Z-stack images acquisition. The developed three-dimensional image processing algorithm uses Z-stack images to segment embryo and micropipette contours and compute their corresponding locations on the image plane. The sliding mode controller based on image based visual servoing concept uses these image plane coordinates to compute an error signal for processing. The velocity output controls the corresponding motorized stage to manipulate the micropipette to penetrate embryo membrane and reach the target location of a selected blastomere centroid. The experimental result with 2-blastomere mouse embryo has shown acceptable accuracy for certain types of cell surgery procedures and potential to extend to 4-blastomere embryo applications.