Xuyang Cao

Senior Algorithm Engineer, Ph.D.

JD.com, Beijing, China
No. 18 Kechuang 11 Street, Tongzhou District, Beijing

[github] [google scholar]

• Built the open-source medical MLLM framework, CitrusV, a unified medical multimodal model that achieves state-of-the-art performance, serving as the foundational model for downstream medical applications.
• CitrusV is the first model in the industry to support both medical image understanding and detection/segmentation, enabling unified parsing of textual, imaging, laboratory, and other multimodal medical data.
• For the first time, addressed a longstanding question by demonstrating that a unified MLLM can surpass expert segmentation models, improving average performance by 112%.
• Introduced chain-of-thought (CoT) reasoning aligned with clinical diagnostic workflows, delivering multimodal reasoning that is explainable, verifiable, and localizable for medical decision support.
• Achieved state-of-the-art performance across 7 core task categories and 21 benchmark datasets. Core achievements released at the annual JDD 2025 (link), forming open-source assets (Code, Technical Report).

• Developed an audio-driven talking face generation framework, to support offline and online medical applications (DAU over 100k), including medical education, medical consultation, etc.
• Proposed JoyVASA, a diffusion-based talking face generation model, which separates dynamic expressions from static 3D identity cues in facial representation, trained a diffusion transformer to generate identity-independent motion sequences directly from audio, and used a dedicated renderer to synthesize high-quality animations.
• JoyVASA achieves real-time performance, with a RTF of 0.76 and a speed of 40x faster than existing diffusion based models.
• The open source projects [JoyVASA and JoyHallo] achieved 1,300+ GitHub stars.

• Served as a key researcher in JD Health Research, focusing on brain science related research, including Brain-Computer Interface (BCI), and the early diagnosis of Alzheimer’s (AD) and Bipolar Disorder (BD) based on medical imaging.
• Research interests include: explainable AI, medical multimodal fusion model, BCI based intelligent rehabilitation, etc.
• Several papers and patents are published, including multiple studies in brain imaging and EEG recognition achieving state-of-the-art results.
• A national key research and development program of China is approved. (project: Virtual Reality Cognitive Rehabilitation Training System for the Elderly).

• Research on deep learning based semantic segmentation algorithms on breast ultrasound images as well as lung MRI images.
• Delved deep into semantic segmentation algorithms, such as fully supervised learning, semi-supervised learning, neural network architecture search, etc.
• Proposed lightweight dilated densely connected network for 3D breast tumor segmentation. The performance improved over 5% compared with classical segmentation networks, while network parameters were over 20 times smaller than classical networks.
• Designed an uncertainty-aware temporal-ensembling model for semi-supervised segmentation. The semi-supervised method achieved 94.4% of the performance of supervised segmentation with only 1.1% labeled data.
• Suggested an NAS-based 3D medical image segmentation framework, and achieved an improvement of 4.2% compared with the state-of-the-art human-designed segmentation network.
• Related journal and conference papers have been published, total impact factor 30+.

• As a core developer, designed an automatic detection and quantitative analysis solution for surface pitting of coupling gear components in train sets, based on computer vision technology.
• Collected coupling gear surface images and applied image recognition algorithms to generate statistics, reports, and alerts for gears exceeding the defect threshold.
• Led the development of algorithms for detecting and segmenting gear surface wear areas, as well as the design and implementation of the software system on the Windows platform.
• Completed defect detection of 104 gear surfaces (internal and external) with defects larger than 1mm within 3 minutes; published related papers and patents.

• Provided a solution for measuring the geometric parameters of an overhead line system using scale factors and frame differences.
• I was responsible for the preliminary algorithm simulation and participated in the hardware structure design work.
• Two related patents have been granted.