It seems most of the article can be summarized by "don't hire inexperienced ML practitioners, they have no idea how to improve a badly performing model", followed by "here are some really simple things a somewhat seasoned ML practitioner would do".