My research here focuses on learning deep probabilistic models, exploiting them for exact and tractable inference, and ultimately aiming to automate machine learning, starting from density estimation for heterogeneous data. I am also interested in integrating symbolic and relational representations into statistical models.