Pretty weird score on test and train set

Can someone explain to me how to interpret this behavior. Is the model overfitting and underfitting at each epoch ?
I work with a dataset of size ~60