Cardiovascular diseases are major challenges to healthcare systems around the world, requiring strong predictive models for early identification and management. This paper introduces a novel model for heart health assessment using an adaptive neuro-fuzzy inference system with a two-stage approach based on the Cleveland Clinic Heart Disease Dataset. In the first stage, nine vital attributes are used for heart disease risk evaluation. In stage 2, this will be further calculated for evaluating heart attack risks from smoking habit and family history. It makes use of Sugeno-type FIS with membership functions Gaussian for stage 1 and generalized Bell for stage 2. Hybrid learning approach combining gradient descent and least-squares estimation maximizes performance. High accuracy is achieved in forecasting conditions such as coronary heart disease, myocardial infarction and aortic aneurysm. The dual-stage model enhances the diagnostic accuracy as well as comprehensive risk assessment that will enable targeted preventive strategies in clinical settings. Using the ‘ANFIS’ package in Python, this framework provides more effective cardiovascular risk evaluation by enabling rapid, accurate detection.