For further inspiration, this is a pretty good-looking "dplyr for Python": https://github.com/machow/siuba

There's precedent in Ecto for a "magic" operator (`^` if I'm not mistaken), so it wouldn't be a stretch to implement it here as well.