Are you a data scientist or machine learning engineer who also has an interest in Elixir? Whether or not you’re using Elixir for data science right now doesn’t matter.
I’m seeking to rally the troops, so to speak. To exchange ideas and experiences.
If you’re interested, please leave a reply on this thread. Maybe share a little bit about what kind of work you’re doing, whether you’re using Elixir, or areas where you would like to use Elixir.
Also, maybe share some of your thoughts on where you think Elixir might fit into your data science/machine learning workflow and where it would not.
I am not your exact target audience because I would like to become a data scientist but I am not yet one.
What would be very interesting to me is: can Elixir be used to supervise and coordinate a number of scientific runtime tasks? We all know the BEAM is not ideal for number crunching and that’s okay; but between Ports, distribution libraries and the Erlang’s node protocol I believe it’s not hard to have tasks started / paused / stopped on demand if the physical / VM nodes implement the proper ways to interact with a running BEAM instance.
Do you have any experience with such a scenario?
What would probably be also very interesting is to research whether the runtimes of languages like R and Julia can freely talk with any BEAM instance (with or without a thin integration layer). They are more and more used in the scientific community and Julia has some pretty interesting things going on – like GPU computations – lately.
If you’re interested at all then you’re definitely in my target audience
I’m lucky to work for a company which values their employees and gives them opportunities to learn new things. As a programmer gone data scientist who has only been in my new role for a couple of weeks, I’m pretty much a newbie to the field of machine learning myself.
Yes! I have similar thoughts. Elixir would obviously be a poor choice for the computationally heavy stuff, like crunching numbers and such. However, it could be the perfect tool for data harvesting and wrangling, federated learning infrastructures and other facilitative aspects.
Many data scientists have adopted Scala and Hadoop to deal with those things.
No, not yet, but I’m currently investigating different methods for serving machine learning models to other systems in our production environment together with my teammates.
I suppose there are some possibilities of interfacing with the BEAM through Erlang or Elixir, by using NIFs, I/O ports or the Erlang distribution protocol. I would prefer to use Elixir if possible.
