Bayesian kernel machine regression (BKMR) is introduced as a new method for estimating the health effects of multi-pollutant mixtures. This approach models the health outcome as a smooth function of the mixture components using a kernel function, allowing for flexible exposure-response functions. A hierarchical variable selection method is incorporated to identify important mixture components and account for their correlated structure. Simulation studies demonstrate BKMR's effectiveness in estimating exposure-response functions and identifying responsible mixture components. The method is applied to environmental health datasets, including a study on metal mixtures and neurodevelopment in Bangladesh, and a toxicology study on air pollution mixtures and hemodynamics. BKMR outperforms frequentist methods in variable selection and captures uncertainty more accurately. The hierarchical approach improves identification of important components in highly correlated mixtures. The study highlights BKMR's ability to handle complex exposure-response relationships and its potential for future applications in environmental health research.Bayesian kernel machine regression (BKMR) is introduced as a new method for estimating the health effects of multi-pollutant mixtures. This approach models the health outcome as a smooth function of the mixture components using a kernel function, allowing for flexible exposure-response functions. A hierarchical variable selection method is incorporated to identify important mixture components and account for their correlated structure. Simulation studies demonstrate BKMR's effectiveness in estimating exposure-response functions and identifying responsible mixture components. The method is applied to environmental health datasets, including a study on metal mixtures and neurodevelopment in Bangladesh, and a toxicology study on air pollution mixtures and hemodynamics. BKMR outperforms frequentist methods in variable selection and captures uncertainty more accurately. The hierarchical approach improves identification of important components in highly correlated mixtures. The study highlights BKMR's ability to handle complex exposure-response relationships and its potential for future applications in environmental health research.