I've spent a lot of time on a automated trading side project of mine but haven't found the strategy yet.
In retrospect time has been spent in completely the wrong areas ( setup a solid backrest platform first, duh )
Backtesting is great for validating ideas initially. Especially, to see if it holds up through abnormal markets like '08, or the volpocalypse, or the recent crash. Watch out for curve fitting, though.
Do you apply ml techniques? I don't suppose you could point a little in a good direction to follow
Something like using NLP on SEC filings the second they come out to catch an initial jump in the underlying would be cool to try out.
Warning: it is a lot of fun, thrilling, but hard to make money. My winning algos took a while to research, longer to automate, and often lost alpha quickly.
Tech stack: I love Ruby, but use Python and a custom language that interfaces with the broker I'm using for execution. More and more decent broker API's are starting to pop up out there. So, you can likely use whatever you're most comfortable with. Alpaca has a decent web API, for example.
I honestly wouldn't recommend getting into it, though, unless you really enjoy geeking out over this stuff and have a decent breadth of knowledge to find your edge. You're also competing against some brilliant PhD types who are just as obsessed and hard working as you. My stuff works in part because it's taking advantage of some things that are too small for the big boys to pay attention to.