Practical applications of machine learning

Hey, I'm finding myself thinking a lot more about how I structure machine learning problems than I do actually playing with parameters and so on. A lot of my programming involves:

• Reading from my database
• Reading from my database
• Reading from my database (it happens a few times!)
• Creating new parameters from combinations of others
• Making bags of words

After all of this I do some ML, and then I spend a while on:

• Checking things worked
• Cross validation + evaluation
• Sometimes "undoing" the work I did earlier, eg, bag of words -> words

So, does anyone have any examples of really well architected applications of machine learning that I could look at?