You may also try the Asymmetric Laplace that was recently added to Pymc 3.11.0: Asymmetric Laplace distribution - Wikipedia. This distribution works by having two exponentials on either side, keeping the same peak but having different sigmas. Would that achieve what you need?