I've been using Julia since 2015 and like it a lot, primarily for two reasons:
the types + multiple dispatch paradigm makes more sense to me than object-oriented programming, more generally I like the syntax,
it's interpreted which helps to try out ideas quickly and yet programs can be very fast.
A list of some of the packages I contribute(d) to. I may write notes on some of them in the future.
MLJ.jl (with Anthony Blaom and a team at the Alan Turing Institute), a Machine Learning toolbox (under active development).
CovarianceEstimation.jl (with Mateusz Baran), a package for robust covariance estimation (actively maintained).
Franklin.jl, an experimental static site generator which I use for this website that allows using LaTeX-like syntax (actively under development).
PDSampler.jl, a package I wrote while at the Alan Turing Institute for piecewise deterministic Monte Carlo samplers and accompanying our paper on using PDMC on restricted domains (infrequently maintained).