I am currently a software engineer at Lyft. I received my PhD in Computer Science from the University of Illinois at Urbana-Champaign, working with Eyal Amir and Roxana Girju on detecting deception in written natural language dialogues.
My research is at the intersection of knowledge representation, natural language processing, and machine learning. More specifically, I am interested in creating computational methods that automatically learn and reason about intentionality, beliefs about beliefs, and belief change in dialogues. My focus is on dialogues where agents with conflicting goals interact to effect each other's decisions. Examples include court trials and negotiations.
My goal is to use these methods to answer queries about the participants' beliefs and intentions, as well as predict their future decisions, dialogue contributions, and actions. Models that can be used for learning and reasoning in this scenario should be flexible, easy to adapt to new domains, and enable deep, semantic understanding.
My work introduces such flexible, knowledge-rich models and computational methods for learning and reasoning with them. The models combine situation calculus, belief filtering, and probabilities, in a framework that captures how agents reason about other agents' beliefs, how they effect and perceive belief change, and how they, in turn, change their beliefs.
I worked on defining or analyzing several logical formalisms: proved reachability for a class of counter automata (octagonal constraints), defined an extended modal logic as an institution, and formalized a probabilistic version of a qualitative spatial logic (Region Connection Calculus).
I received my BS in Computer Science from the Polytechnic University of Bucharest and MS in Informatics from SNSB, working with Till Mossakowski