Barrier pattern to deal with threads, although I mostly can let libraries deal with that.
Those are two I can remember using recently. I once worked on a very big graph search problem using A* and some other algorithms. That was fun.
I haven't implemented a single computer science-style algorithm since I graduated. Even the rare times I need one there are perfect (i.e. bug free/edge case free) pre-existing examples available.
This is likely true for the silent majority who are doing CRUD or web development.
Another experimental algo which compares rich vectors for similarity which is a variant of an uncentered Pearson Correlation
Since most algorithms apply to very specific data and needs of that data to do something.