Work Experience

From August 2023, I work as a postdoc research associate at the University of Edinburgh, School of Informatics on Large Language Models and Reinforcement Learning for robotics.

From 2022 to 2023, I worked as a full-time Applied Scientist II at Amazon on developing machine learning recommender systems to provide personalized recommendations to customers.

During this internship, I studied the properties of game-theoretic payoff allocation methods in submodular multiagent machine learning (ML) applications. Payoff allocation is a key question in multiagent systems, i.e., how to evaluate the importance of each entity towards the collective objective? However, classic literature on cooperative game theoretic solution concepts mostly focus on supermodular games. In particular, an important notion that arises in the submodular applications is redundancy, due to abundant data or malicious replicating manipulations. In this project, I studied the properties and robustness of game-theoretic solution concepts such as the Shapley value against replication manipulation in submodular games. And the theoretical findings are empirically evaluated on an emerging submodular ML application, i.e., the ML data market.

This is a Natural Language Processing (NLP) internship. During this internship, I worked on dialogue systems, and successfully designed and developed a machine learning user simulator using state-of-the-art NLP techniques such as semantically-conditioned LSTM's. The user simulator can simulate the human behaviour, and achieve multi-attribute goals by carrying out multi-turn dialogues with a dialogue system like Siri in natural language.