In this study, we use smartphones to examine the prospect of automating the identification of malaria parasites amid heavy stains of blood. We developed the deep learning method which is first technique (superset of machine learning) capable of recognizing malarial parasite in thick blood stain images on mobile devices. Below is how our two-stage processing works. To begin, we use a Recursive Global Minimum Screen (IGMS) technique that rapidly screens a thick smear picture for parasite candidates based on intensity. Finally, each potential candidate is assigned to the parasite or background category employing an established Deep Neural Network (CNN). With this publication, the academic community now has access to 1819 images of thick smears or stains taken from 150 patients. In this research, we demonstrate how we fed this data set into our deep learning framework for training and testing purposes. Results: In a cross- evaluation which is five-folded at the patient level, the tailored CNN model performed exceptionally well in terms of accuracy (93.46percent ) of the respondents 0.32%), region under the curve (98.39percent ) of the respondents 0.18%), sensitivity (92.59percent ) of the respondents 1.27%), specificity (94.33percent ) of the respondents 1.25%), precision (94.25percent ) of the respondents 1.13%), and a negative predictive value (92.74percent ) of the respondents 1.09%). At both the image and patient levels, the automatic parasite identification and ground truth exhibited high correlation coefficients (>0.98), demonstrating the system's effectiveness.