About Me
Experienced AI Research Scientist with expertise in foundation models, autonomous driving, natural language processing, and reinforcement learning.
Proven track record of leading research teams and developing cutting-edge AI technologies with significant commercial impact.
I am currently interested in developing foundation models that enable lifelong learning agents.
Work Experience
- Proposed next-generation in-context learning and self-evolving foundation decision models and world models
- Small-scale model validation for cross-scenario navigation, cross-embodiment motion control, etc.
- Main results available at: OmniRL Project Page
- Proposed and deployed multi-objective Pareto optimization DPO solution, significantly improving 10-billion-scale model performance
- Explored autonomous evolution mechanism integrating model generation and evaluation, achieving breakthroughs in multi-round DPO performance
- Explored multimodal and embodied data synthesis directions such as autonomous driving
- Assisted the Group Technology Strategy Committee in developing next-generation autonomous driving technology plans
- Proposed end-to-end autonomous driving models integrating world models and decision models
- Real-vehicle deployment testing, achieved 98% success rate at Haidian test track with random disturbances
- General Dialogue Group: Led release of PLATO, China's first commercial generative dialogue foundation model
- Established AI bio-computing direction, open-sourced PaddleHelix platform, published in Nature sub-journals
- Led reinforcement learning direction, open-sourced PARL framework, won NeurIPS competitions 3 times
- Collaborated with Baidu Maps, improved ETA accuracy with graph models
