Bangladesh is an extremely flood-prone country due to its geographical location at the downstream end of the Ganges, Brahmaputra and Meghna (GBM) river basin. Flood destroys agricultural products of large areas and causes loss of lives and damage to infrastructures. Heavy rainfall during the monsoon season is the major cause of flooding in this region which occurs almost every year. However, the lack of observations of rainfall in the upper catchment areas outside Bangladesh makes flood forecasting challenging in this region. In addition, errors in rainfall forecasts and lack of high-resolution bathymetry and topographic data put major constraints to flood forecasting in Bangladesh through hydrologic and hydrodynamic models. Currently Flood Forecasting and Warning Centre (FFWC) of Bangladesh Water Development Board (BWDB) is producing short-range flood forecasts with a lead time of up to three days. However, medium-range (3 to 5 days) forecasts are crucial for reducing flood-related losses as they provide more time for decision making and preparation compared to short-range forecasts. In this study, a flood forecast model based on Artificial Neural Network (ANN) has been developed for the Kushiyara river which is one of the major rivers of the northeastern region of Bangladesh. Rainfall data from the fifth generation European Centre for Medium-Range Weather Forecasts Reanalysis (ERA5), daily Terrestrial Water Storage (TWS) from the Global Land Data Assimilation System with the Gravity Recovery and Climate Experiment Data Assimilation (GRACE-DA) and daily Surface Soil Moisture data from Soil Moisture Active Passive (SMAP) have been used as input to the model. The model shows reasonable accuracy in forecasting the water level of the Kushiyara river at Sheola station with a lead time of up to seven days. For 1-day lead time, the correlation coefficient (R) between the observed and simulated water levels is 0.97. The performance of the model is also promising for a medium-range forecast (R=0.93 for 7-day lead time). This study indicates that the release of daily GRACE gravity field solutions in near-real-time may enable us to forecast and monitor high volume flood events in this region.