Microclimate variables like temperature, humidity are sensitive to land surface properties and land-atmosphere connections. They can vary over short distances and even between sections of the farm. Getting the accurate microclimate around the crop canopy allows farmers to effectively manage crop growth. However, most of the weather forecast services available to farmers globally, either by the meteorological department or universities or some weather app,  provide weather forecasts for larger areas. To address this issue we developed a ~100 m spatial resolution AutoML framework that predicts hourly temperature and humidity over a period of 24 hours. The system uses one year of historical data from both IoT sensors and local weather forecasts for training and predicts temperature and humidity using individual models. The models were developed using a gradient boosting machine (GBM) approach. To account for model drift and data drift, an autoML framework was developed to automate model training on a monthly basis. The autoML framework uses i) a Bayesian optimization-based Hyperopt library to automate hyperparameter selection for the GBM models and ii) MLflow for model training, logging, and deployment purposes. For continuous deployment of the models, the autoML framework was integrated with Kubeflow for production-level serving. The models were developed using historical data from 6 different districts of Maharashtra, India and the accuracy was tested for over 50 grape farms (~50 ha) from that region via live deployment. Compared to local weather forecasts, the models showed a 15% and 30% decrease in mean error for temperature and humidity prediction, respectively. The developed microclimate framework also outperformed in predicting extreme temperature and humidity conditions by ~30%. Timely prediction of extreme weather conditions would be helpful in effective crop protection and crop management. Though the AutoML Microclimate framework was developed and tested for grape farms in the Maharashtra region of India, it can be easily extended to other regions and crops as well.