Agriculture is one of the main pillars of food pro- duction in the world, which is experiencing increasing problems because of the variability of climatic conditions, poor use of available resources and decreasing soil fertility. Conventional farming practices are prone to manual surveillance and judgment which are inaccurate, labour-intense practices incapable of keep- ing up with the dynamic environmental factors. Crop monitoring and yield prediction have been implemented using the existing methods like rule-based systems and classical machine learning like Decision Trees and Random Forests. These models however do not usually account for complex nonlinear relationships and temporal dynamics of the environmental and soil data to make optimum decisions and to be scalable. In order to address these shortcomings, the current research suggests sustainable crop growth optimization system based on Artificial Intelligence and a CNN BiLSTM deep learning model. Convolutional Neural Networks (CNN) are used in the model to obtain spatial relation- ships between features, including soil moisture, pH, temperature, humidity, and light intensity, and the Bidirectional Long Short- Term Memory (BiLSTM) element is used to obtain forward and backward temporal relations among sensor data sequences. The dataset, which will be used, is Smart Agriculture and Plant Health Monitoring using IoT, which offers multivari- ate environmental measurements. Experimentally, it is shown that CNN-BiLSTM model works better in prediction accuracy, temporal stability and generalization to achieve considerably higher improvements in root mean square error (RMSE) and mean absolute error (MAE). The suggested model is effective in predicting the best irrigation and environment changes to ensure the sustainable management of the resources, reduction of waste of water and other fertilizer, and increase of crop production and ecological stability. Such a strategy opens the path to smart precision farming and data-driven sustainable farming.