Rocket Scientist, Quant, candlestick maker, have you thought about Shapes and Motifs as the next big idea?  Could be, mainly how do you discern patterns in really large datasets?  Take a look at: http://msrvideo.vo.msecnd.net/rmcvideos/103425/dl/103425.pdf for an interesting take on some thinking on how to deal with extremely large datasets.  No algorithms though, which definitely sucks.  So I found a paper at that discusses this in more depth:

http://www.springerlink.com/content/v4p5747q2m665385/ (Make sure to download the PDF, as the online version is a little hard to read)

So let’s say you need to create a new system to perform classification of certain datasets, let’s say images of oranges being sorted during harvest, you are reviewing the images for orange related diseases.  Then you might consider the use of Shaplets and Motifs, but right now I am not quite certain how, the why or motive seems obvious it is a safer algorithm in many cases.  In the case of Azure, you would generate a process, algorithm and then offer your algorithm as a subscription, if users needed more processing power then they would pay for the extra CPU utilization, and you would charge an appropriate mark-up.

So take a look at these papers, I have a feeling that I am going to put together a series on how to use Shaplets and Motifs with your code.  Now this is something you could use as a Quant

image