The "How I did it"[0] section of Kaggle's "No Free Hunch" blog is really great for getting practical insights into solving machine learning problems. All the posts are short, and unless you're an expert in ML, will likely give you leads all a lot of new material to learn. The pragmatic bent is what really makes it such an excellent resource, there's a huge gap between the mathematical foundations of ML and the solving real world problems side of it.

0. http://blog.kaggle.com/category/how-i-did-it/