● Heatwave Watch System, a collaborative R&D product with India Meteorological Department, Pune to watch heat and predict probable heatwaves to issue early warnings for the public to take precautionary measures. ● It is an outcome of extensive research, data engineering, models' training, and exhaustive performance evaluation. ● Data Source: The data for this research is provided by IMD, Pune. The system is developed using historical data of past 71 yrs – from 1951 to 2021. As the data has a temporal component, 80% data (of yrs 1951 to 2010) used for training and the recent 20% data (of yrs 2011 to 2021) used for validation. The dataset dimensions are 31×31, and consists of attributes Timestamp, Latitude, Longitude, and Max Temperature, which allows learning temperature trends over different seasons, regions, and time. ● Data Pre-processing: The data is extracted from the web-based repository with GRD files of maximum observed temperature for spatio-temporal analysis. Further, the data is fragmented into the 07 IMD-defined climatic regions of India & 04 climatic seasons. The data is further pre-processed to handle noisy/missing values and normalization, thus assuring stable input to the AI/ML models. ● Machine Learning: For learning spatio-temporal patterns of temperature, the work uses Long Short-Term Memory (LSTM) networks, which are considered to be the most effective, especially for weather prediction. From the regional segments of data, 28 lightweight data-driven LSTM models are trained. The results from these models are further recombined and rescaled to produce the 31 × 31 dimensional spatio-temporal predictions. The models are capable to predict maximum temperature values, and assist heat watch. ● Prediction: The work uses maximum daily temperature readings of past 14 days to produce maximum temperature predictions for leads of next 7 days. ● Validation: The current accuracy of predictions is 92.10%. Further validation of this research is in-process.