
Below is a short summary of my work and research interests. More information is in my resume.
I have worked throughout my career predominantly on NLP problems; mostly on conversational systems, representation learning, and more recently on Large Language Models fine-tuning and evaluation. I built, led and mentored ML teams and helped several startups build their first AI products.
Research
My research falls in the intersection of AI, NLP and social psychology. Since our subjective experience of life is deeply grounded in the quality of interactions with those around us, I am studying how AI systems can improve both human-human and human-computer interactions. Ultimately, my goal is to build “social-machines” that are truly suitable for human interactions. Through demonstrating social reasoning and emotional abilities, such systems could serve valuable roles in our daily lives as well as serve as tools for explaining social phenomena. I'm currently focusing on improving theory of mind abilities of LLMs using Reinforcement Learning and conversational data as part of my part-time PhD at the Oxford Internet Institute under Dr Scott A. Hale's supervision. During my masters at University of Maryland, College Park I worked with jordan boyd-graber.
Contact
Email: [mkhooja at gmail dot com]
LinkedIn: [https://www.linkedin.com/in/khouja/]
Google: [Google Scholar Page]