• Expertise: Large-Scale AI Systems, Distributed Training, LLM Post-training.
• Mission: Building the instruments to explore the nature of intelligence.
• Ex-ByteDance Research Engineer: Architected the core RL training system for the Doubao foundation model.
• Creator of VeRL/HybridFlow: From a research paper HybridFlow to a popular open-source project VeRL.

📫 Open to collaborations and conversations: [email protected]

人猿相揖别. Apes and humans part ways.

Hi. I'm 叶子凌锋/Ye Zilingfeng/Enoch Yip.

My journey began in physics, driven by a desire to map the material world. Yet, the greatest territory—the nature of intelligence and consciousness itself—remained uncharted. This fundamental question guided my path from the study of matter to the engineering of mind, grounding my work in the practical challenges of distributed systems and large-scale AI infrastructure.

I used to be a research engineer @ByteDance, Seed, where I led the architecture and implementation of the company's first large-scale RL system based on Ray to exclusively power the training of all production releases of the Seed foundation model (a.k.a Doubao).

This system became a research paper HybridFlow and was turned open-source as VeRL, gaining significant traction and acclaim within the community.

Building this system from concept to a cornerstone of production was an incredible journey, and it clarified my path forward. In early 2025, I chose to leave ByteDance, driven by a desire to apply these experiences in a more agile and pioneering environment.

I believe we are living in a pivotal moment. For the first time in human history, LLMs allow us to turn the lens of science inward, to study intelligence as an objective phenomenon. My work is to build the very instruments for this exploration. Grounded in a deep respect for science, I am motivated by a belief in the ultimate power of intelligence: its capacity for self-reference, which will enable it to endlessly understand and transcend itself.