My research journey began with temporal point processes—especially Hawkes processes (hawkes-knowledge), which model bursts of activity over time. This foundational work led me to a broader challenge that drives my current agenda: how do we move from noisy, discrete behavioral data to reliable models, predictions, and decisions? I study this challenge within complex social and technical systems where actions unfold over time, interact through networks, and reflect strategic incentives. In online information ecosystems and financial markets, actors post, trade, coordinate, and adapt to platform interventions. Analyzing this observable microstructure of behavior allows us to understand information flows, detect harmful or coordinated actors, and design more effective policies. Methodologically, my work bridges artificial intelligence, reinforcement learning, network science, and operations research to address three core pillars: inference (what model best describes the data?), prediction (how well can we anticipate the future?), and optimization (what is the best decision?). A central theme of my research is end-to-end learning—building scalable, constraint-aware frameworks that integrate these pillars into a single pipeline. More recently, I am exploring how large language models (LLMs) and AI agents are fundamentally altering social systems and platform dynamics. My ultimate goal is to build decision models that are theoretically rigorous yet practically deployable. If you are interested in collaborating or exploring project opportunities, I’d be glad to connect via email!