I currently am a research scientist at Google. Here is a Google research blog post about one of the systems I co-designed.
During the four years that I was at Stanford I worked on a couple of related topics:
Active Learning. The standard framework in Machine Learning presents the learner with a randomly sampled data set. There has been growing interest in the area of Active Learning. Here, one allows the learner the flexibility to choose the data points that it feels are most relevant for learning a particular task. One analogy is that a standard passive learner is a student that sits and listens to a teacher while an active learner is a student that asks the teacher questions, listens to the answers and asks further questions based upon the teacher's response. I investigated techniques for performing active learning in three widely applicable situations: classification, density estimation and causal discovery. Our results showed that active learners using these techniques can outperfom regular passive learners substantially - particularly in the text classification and image retrieval domains (very relevant domains given the recent explosion of readily available data from the internet).
Restricted Bayes Classifiers. Support Vector Classifiers are a core technology in modern machine learning. They have strong theoretical justifications and have shown empirical successes. My research led me to attempt to cast them in the probabilistic framework. The technique for achieving this appears applicable to a wider range of classifiers.