The little bit I've done with Ruby on Rails was an opportunity to delve into HTML as a spec/standard/language. I love HTML as a medium for written expression—tend to be quite digressive and referential in my writing and I love how I can drop clues or actual links to my references without having to interrupt the flow.
This site is actually an opportunity for me to dip a toe back into these waters, though I must confess that I am cheating a bit by using the bootstrap framework.
This looks to be a perfect culmination of HTML & other web standards—a means for bringing what would otherwise be static data visualizations alive and enabling the viewer to make their own discoveries about the data. Plus they can be distributed as single html files! No more cramming everything into stultifying rich text or portable document format. I am in.
I've lately been making a concerted effort to learn this, and have started a project using D3 to document HCSRN's Virtual Data Warehouse--implementations and specs. Have a look.
Working on the pathology report text classifier we used on a study to do rapid cancer case ascertainment was an absolute blast, and whet my appetite for getting further in to that. More recently I've used ML to predict impending hospitalizations in Kaiser patients (improving on the predictions offered by the ACG software even!) which has been huge fun.
In the last few months I've taken on a python project (the aforementioned ML prediction of hospitalization, for which I used scikit-learn) and have quite liked it. It's not ruby (sigh) but it's fast and there's such a community built up around it.