Always predicting all data to one class

Classification task
I am using a data of Age, tenure and Salary to predict turnover in which there are 2999 rows in total and 1836 employees left the company and 1163 not left the company.

Now i have built a deep learning model for it and able to run it successfully utilizing only 2 dense layers but for a strange reason the model is always predicting single class only in all cases whether testing/validating or any other data is used.

What is wrong with DLS ?

Please let us know what is the accuracy that you have achieved with this model ?

The accuracy is about 64% percent. I can also share a csv file with you.