Accurately predicting temperature weeks in advance can help communities better prepare for extreme heat, especially in areas that regularly experience very high temperatures. This study focuses on forecasting summer temperatures in the hottest parts of India, where average temperatures during April to June often exceed 38 °C. Using machine learning (ML) methods, developed models to predict daily maximum temperatures 15 to 60 days ahead using meteorological data from 1990 to 2019. Two advanced ML models-SVR and XGBoost were tested. The results showed that SVR worked better for shorter forecasts (15–30 days), while XGBoost gave more accurate predictions for longer periods (30–60 days). Although forecast accuracy varied depending on weather conditions and the data used, the models were able to capture major temperature patterns. These findings show that ML models can support early warnings and better planning for heatwaves, helping to reduce risks to health, farming, and energy systems in India's most heat-prone regions.