Both are regularization techniques, but in lasso the lasso regression coefficient is |beta| which leads to making some of the coefficient estimate exactly equal to zero and hence can also be used as feature selection. While in ridge regression, the penalty causes the coefficient estimate tend towards zero and hence does not remove any feature completely. Lasso is used when a relatively small number of features have substantial coefficients and others close to zero. Ridge is used when coefficient estimates of all the features are approximately of similar size.