Human exposure to extreme heat and cold poses increasing risks to public health, labour productivity, and urban sustainability, particularly in densely populated and climate-sensitive regions such as India. Human-perceived temperature (HPT) indices provide a more realistic measure of thermal stress than air temperature alone by integrating multiple meteorological factors. Here, we present the Human Thermal Index Collection for India (HiTIC-India), a high-resolution daily gridded dataset comprising twelve widely used HPT indices at 1 km spatial resolution for 2003–2020. The indices are initially derived from ERA5-based meteorological data and then downscaled using a Light Gradient Boosting Machine (LightGBM) framework. This downscaling incorporates satellite-derived land surface temperature, precipitable water vapour, population density, and topographic variables (slope, elevation and aspect) to generate spatially continuous predictions at 1 km resolution. Model valuation shows high prediction accuracy across all indices, with a mean root-mean-square error (RMSE) of 3.12 °C, a coefficient of determination (R²) of 0.89, and a mean absolute error (MAE) of 2.39 °C. The resulting dataset significantly captures local-scale variability in heat and cold stress across India’s diverse climatic and physiographic zones. HiTIC-India also supports numerous applications, including public health risk evaluation, urban heat exposure analysis, labour productivity assessment, and climate adaptation and mitigation planning. By providing consistent daily HPT datasets, HiTIC-India provides a comprehensive, high-resolution, and publicly accessible resource for climate–health research and evidence-based decision-making under warming climate.