Thanks! I wonder: working as data scientist involves a lot of communication (regarding explaining results and discussing what to focus on next, etc.). By what channel do You communicate? Also do You sometimes travel on-site?
That's a great question. I'll definitely add it to my list of interview questions I'll ask other folks.
As for me communication is:
1) Lots of time on Slack (a IRC-like tool) where we have channels set up for each of the DS projects and for the DS team. We can reference work tickets, paste code and results and have a logged/archived conversation. This is better than a standard chat program in that regard, and it makes brining in others into a conversation easy.
2) Lots of video chats. Hangouts is buggy and CPU intensive, but easily allows you to share terminal windows - where I do a lot of work in AWS EMR. Sometimes I've spent the better part of a day co-working with someone via a Hangout.
3) For simple-ish data Google Sheets works out very well for sharing data. I've become pretty adept at creating the graphs I need. It's easy to drop them into a presentation. Complex data usually means you need to work harder to make it more simple. Either that or you should be actively collaborating on it.
4) I always log all my work either via Git or a running presentation (Google Presentation or Power Point) of notes and images - an Jupyter / IPython Notebook would work too, but whatever suites your style. We routiunely present analysis or results to colleges remotely via a presentation like this. Google Presentation is convenient because you can see which sheet everyone is looking at at a glance.
5) We manage our current and future work via JIRA. We've also tried Blossom, Basecamp, Sprintly and Asana. Any of those can probably work fine. Whenever we have a research idea we make a card and put it in the queue. We track what everyone is working on on a given day by looking at the "today" stack of cards. Cards reference code, plots and links to analysis presentations.
6) When coding on an algorithm or problem with a coworker we use Madeye.io or Floobits. Other coworkers use Screenhero (but that's more like pair programming which I don't like) or the collaboration plugins for Atom or SublimeText.
What I really really miss is the ability to gather around a whiteboard and work through some analysis. I haven't found a good solution to this yet. I've purchased a cheap drawing tablet from Amazon but I haven't found adequate software to let me draw with a coworker in realtime.
I haven't had to do any on-site travel up till now. That may change in the future, but for now I resist it as much as possible. Our office has a East and West coast location and 7-8 totally remote workers. The goal is to make the "on site" notion obsolete. Co-located workers shouldn't be 1st class.
I work remotely as a data scientist and I can't see myself ever going back. I've been blogging jobs I find for remote DS work. Hope to also add some interviews with other remote DS folks in the future.
