Hooray~ you caught me! I am Junhao Hu, a fifth-year Ph.D. student at School of Computer Science, Peking University, advised by Prof. Tao Xie.

• Undergraduate: I graduated from Kuang Yaming Honors School, Nanjing University in 2022, where I was lucky to be advised by Prof. Yanyan Jiang, known to many students for his popular operating systems course on Bilibili.
• Ph.D.: I have spent most of my Ph.D. in industry, looking for real-world problems to work on. During these five years of internships, I was lucky to work with and be advised by Prof. Chenxi Wang, Yizhou Shan, Hailin Zhang, Kechi Zhang, and Xusheng Chen.
• Now: I am an RSI research intern at Tencent Hunyuan, and may join full-time after graduating in 2027.

My research centers on algorithm–infrastructure co-design for LLMs and agents. I started with Infra4AI, first building inference infrastructure (e.g., EPIC, DeepServe, RedKnot, Hy4, MiMo-V2) and then agent harnesses and memory systems (e.g., PenguinHarness, Akashic). As AI grew more capable than I am at building infrastructure, I chose to build the AI that replaces me rather than compete with it. Therefore, I now focus on AI4Infra, specifically RSI (recursive self-improvement) for agent harnesses: training models to write and evolve their own harnesses (e.g., GDPevo, Awesome-RSI). Next, I may extend AI4Infra to inference and training infrastructure.

For more information, please hang around this website or refer to my CV. Feel free to reach out for discussion or collaboration!