Dongkuan (DK) Xu

I am an Assistant Professor of Computer Science at NC State University, where I lead the Generative Intelligent Computing (GIC) Lab. My research focuses on machine learning, natural language processing, and computer vision, with an emphasis on efficient and reliable AI systems. I have received several awards, including the Microsoft Accelerating Foundation Models Research Award 2024, the NCSU Carla Savage Award 2024, the Best Paper Award Runner-Up at IEEE IPCCC 2024, and the Best Paper Award at ICCCN 2023. I hold my Ph.D. from Penn State, M.S. from the University of Chinese Academy of Sciences, and B.E. from Renmin University of China.

I am also involved in MARINA, a research initiative at the intersection of AI and ocean sciences, focusing on advancing ocean modeling, environmental monitoring, and AI-driven scientific discovery. As part of my broader work on AI systems, I have contributed to the development of Synthora and Gentopia. Synthora is an extensible AI platform designed to enhance the computational efficiency of LLM-driven applications through full-stack software-system hardware optimization. Gentopia is a lightweight framework for LLM-driven autonomous agents, providing essential components for building, testing, and evaluating AI agent applications.

Outside of work, I am a fan of American football (Nittany Lions, Eagles, Cowboys) and enjoy working out and playing soccer.

Email  /  CV (Nov 2024)  /  Twitter  /  Google Scholar  /  LinkedIn   

📣 Seeking 1-2 PhD students, encouraging candidates from underrepresented groups. Interested applicants should email "PhD Fall 2025" with their CV.

News

My research focuses on scalable, efficient, and reliable learning for AI systems, particularly in the context of large language models (LLMs), diffusion models, and other generative AI models. I explore methods that enhance inference-time performance, shift computational efficiency curves, and ensure robust and reliable learning across diverse applications.

• Test-Time Scaling & Adaptation. I study test-time scaling laws to dynamically enhance AI performance at inference and testing stages. My work explores adaptive inference strategies, while also leveraging test-time training techniques to generate and utilize high-quality data. Additionally, I focus on optimizing reasoning workflows to improve model accuracy through self-refinement mechanisms.

• Shifting the Computational Curve. To optimize the efficiency-performance tradeoff in AI systems, my research focuses on accelerating models and enabling dynamic computation. This includes model compression and distillation techniques, as well as conditional computation methods. I also explore system-algorithm co-design for large model inference and training, along with hardware-algorithm co-design to develop energy-efficient neural network architectures and optimize AI-accelerator collaboration.

• Reliable Learning & Robust Generalization. Beyond efficiency, I focus on trustworthy AI by ensuring robustness, uncertainty quantification, and adaptive learning. My research includes uncertainty-aware AI techniques that enable models to recognize and communicate prediction uncertainty, improving their reliability in deployment. Furthermore, I investigate adaptive learning strategies that allow AI systems to refine their knowledge efficiently from new data with minimal supervision.

• Applications. My research contributes to real-world challenges in Education, Agriculture, Networking, Sustainability, Climate, Transportation, Healthcare.

• MISSION: To revolutionize LLM decoding with adaptive, compute-optimal strategies and test-time learning

  ✅ Fine-Tuning-Free Efficiency: Achieve 2.5x speedup without the need for fine-tuning

  ✅ Dynamic Adaptability: Align token predictions with evolving output distributions in real time

• Project Website [link] / AAAI'25 Oral Paper Award

• MISSION: To build artificial general intelligence through collective growth of generative intelligent agents

  ⭐ Demo I (Create Agents), ⭐ Demo II (Customize and Interact with Agents), ⭐ Demo III (Evaluate Agents)

• Project Website [link] / Full Documentation [link] / EMNLP'23 (System Demo Track)

• MISSION: To reshape next-generation education through trustworthy generative AI technologies

• ✅ Innovations: ① Trust & Transparency, ② Dynamic & Controllable, ③ Multimodal & Multifunctional

• Project Website [link] / Video Demo [link] / AAAI'25 (Best Demonstration Award)

*Last updated on 03/17/2025*