In the face of the global COVID-19 pandemic, this study harnesses machine learning techniques to address two pressing challenges: disease diagnosis and drug discovery. Utilizing a dataset sourced from the Indian Ministry of Health, we developed a logistic regression model to predict COVID-19 diagnosis based on eight binary features, achieving an accuracy of 93.5%. Concurrently, we employed graph neural networks (GNN) for drug discovery analysis, yielding a promising hit rate of 72.3%. Compared to other models, our chosen methodologies either matched or surpassed benchmark performances, signifying their reliability and robustness. The high accuracy of the disease prediction model suggests its potential utility in real-world scenarios, especially in prioritizing testing, triaging patients, and optimally allocating healthcare resources during resource-constrained times. The success of the GNN in drug discovery indicates its potential in shortening the drug development pipeline by identifying promising drug candidates more efficiently. While our results are promising, future research should consider integrating more diverse datasets, exploring ensemble machine learning techniques, and validating the identified drug candidates through rigorous laboratory experiments and clinical trials. This research underscores the significant potential of machine learning in navigating health crises and propels us toward a future of more informed, data-driven healthcare interventions.