Databricks implements max_by and min_by, which is similar to keep (dense_rank last order by) :

select id,
       max_by(column_a, column_b)
from table_name
group by id;

Both functions can also be invoked as a window function using the OVER clause.