Shuguang Chen

I'm currently working as a Staff Applied Scientist at DigitalOcean. Previously, I was a postdoctoral researcher at Purdue University working with Prof. Guang Lin. I received my Ph.D. in Computer Science from the University of Houston, advised by Prof. Thamar Solorio at the RiTUAL Lab, where I worked on information extraction.

My research centers on agentic AI, the study of intelligent systems that reason and act reliably in diverse and dynamic environments. Drawing on my background in NLP, I aim to enable machines to reason over complex information, interact effectively with the world, and support trustworthy decision-making in real-world settings. My current work pursues this along three directions:

• Reasoning and Orchestration: how agents decompose multi-step tasks, select the right model for each step, and coordinate across sub-agents.
• Memory and Personalized AI: how agents manage memory across sessions to support future decisions and adapt to individual users, drawing on past interactions and the outcomes of their actions.
• Recursive Self-Improvement: how agents iteratively improve their intelligence through a closed loop of feedback: learning from past experience, simulating new trajectories through varied harness, and optimizing their strategies and execution.