Hi, I spend half of my time on data science consulting in Python, and the other half on research in AI with Julia. Most of my data science work involves making clever use of already-built methods (i.e. scikit-learn), and Python's extensive libraries work great for this purpose. Furthermore, a lot of machine learning can be expressed with linear algebra, and numpy is very decent for writing my own algorithms. For these reasons and for Julia's pre-1.0 status, I don't see myself recommending Julia over Python to a client in at least a year.
That said, as a data scientist, you shouldn't frame yourself as a "Python programmer" or a "Julia programmer". You should make sure that you're good at programming, but your #1 skill will be all the statistics/probabilities/machine learning knowledge and experience that you have. Keep reading, keep practicing. If you really want to use Julia, now is a great time to contribute some random forest implementation to JuliaML. That will look great on your resume. Finally, have you read about the Python Paradox <http://www.paulgraham.com/pypar.html>? That was written 10 years ago. If it was written today, it would be called The Julia Paradox. There might not be a lot of places where you can use Julia, but good data science employers will look favorably on your Julia experience, even if they ask you to use Python/R/Matlab in your job. Good luck, Cédric On Tuesday, December 1, 2015 at 3:19:13 PM UTC-5, [email protected] wrote: > > Hi Everyone, > > > I'm currently learning data science and I have a cs101 with python > background. > > > I have this nagging feeling that Julia is going to be huge (and its > pleasing to code in) and so as I begin to learn stats with python, I keep > drifting over to Julia. Learning python seems like I'm investing in a stock > at its peak and its only downhill from there. > > > However, I also have a nagging feeling that its not ready for productive > data to data analysis or data engineering type production, that the job > prospects will be slim for a while and that I will spend too much time > chasing pycall etc errors. > > > Also with python I get can get a backup sysadmin, backend web etc job if > it turns out I'm terrible at stats. > > > I'm thus vacillating and not sure which language to learn here on out. > > > Are my considerations sound? Any other thoughts on this please? > > > Thanks! >
