For each element, define <code>right[i]</code> as the value of the greatest element with an index greater than <code>i</code>.

Start iterating from the beginning, define a set containing the elements seen so far.

When you are at index <code>i</code>, use binary search on the set to find the greatest element on the left of index <code>i</code> that is smaller than <code>nums[i]</code> and name it <code>greatest_left</code>.

Also check that <code>nums[i] < right[i]</code>.

If the above conditions hold, then <code>ans = max(ans, greatest_left - nums[i] + right[i])</code>.