Malaria, a parasitic infection, is a dangerous public health issue and being endemic in over 80 countries, contributes significantly to mortality. The seriousness of the disease stems from the disruption of vital functions of the blood cells due to its infection by the parasite. It is therefore indispensable for a well-timed and precise diagnosis of malaria, for an efficacious disease control. Some common diagnostic techniques include microscopic analysis of blood smears, antigen detection using RDT and molecular analysis using PCRs. However, the conventional diagnostic techniques are time-consuming and require highly skilled personnel, which is not easy to come across, especially in environments with low resources. This paper aims to overcome these pitfalls, by implementing deep learning architectures. The models explored are Convolutional Neural Network (CNN) models, transfer learning architectures, including DenseNet121, ResNet50, VGG19, and Ensemble voting with pretrained models, wherein the models are tasked with distinguishing between infected and uninfected red blood cell images. Further the feasibility of the models chosen, is gauged using evaluation metrics such as precision, accuracy, recall and F1 score. Ultimately the paper aims to develop an effective and scalable tool to significantly enhance the accessibility of malaria diagnosis, reducing the dependency on specialized technicians and expensive diagnostic equipment.