🚀 Designed parallel inference pipelines for privacy-preserving MLaaS, reducing latency on edge devices
⚡ Achieved ~60% faster inference with overlap scheduling and pipelining techniques
⚙️ collaborated with PyTorch + CUDA kernels, scalable to cloud deployment
☁️ Demonstrated edge-to-cloud efficiency, bridging research and real-world deployment

• Xu, X. et al. SPOT: Structure Patching and Overlap Tweaking for Effective Pipelining in Privacy-Preserving MLaaS with Tiny Clients. 2024 ICDCS.

• Xu, J., Guan, C., & Xu, X. (2018). Energy-efficiency for smartphones using interaction link prediction in mobile cloud computing. CCF Conference on Computer Supported Cooperative Work and Social Computing.

☁️Clear out long-standing rumor in related research

🏗️ Developed quantization-friendly network linearization for secure inference
📉 Reduced communication overhead by ~50% without loss of accuracy
⚙️ Optimized LLM and vision model architectures for edge and cloud deployment
🔐 Enabled model compression with minimal performance trade-offs

• Xu, X. et al (2025). PrivShap: A Finer-granularity Network Linearization Method for Private Inference. TMLR

🔐 Designed secure and communication-efficient transformer inference protocols
📉 Reduced communication cost by ~60% while maintaining model accuracy
🚀 Improved throughput & latency for large-scale privacy-preserving inference
☁️ Applied to LLMaaS and enterprise secure AI platforms, bridging research to deployment

• Xu, X. et al (2024a). Comet: A communication-efficient and performant approximation for private transformer inference. ArXiv Preprint ArXiv:2405.17485.

⚡ Built resilient federated learning frameworks with dynamic trust adaptation and Byzantine robustness
🤝 Developed personalized collaboration mechanisms for heterogeneous clients
🔄 Introduced anti-forgetting strategies for incremental model updates
☁️ Enabled robust and efficient distributed training across cloud and edge environments

• Bai, Y., Wang, Y., Xu, X., Yang, Y., Batool, H., Iqbal, Z., & Xu, J. (2025). AsyncDefender: Dynamic trust adaptation and collaborative defense for Byzantine-robust asynchronous federated learning. Computer Networks, 111430.

• Wang, Y., Xu, J., Yuan, Q., Bai, Y., Yang, Y., Xu, X., & Batool, H. (2025). Fgcfl: a fine-grained clustering framework for federated learning with heterogeneity data. J. Supercomput., 81(14), 1328. 

• Xu, J., Zhao, Y., Li, X., Zhou, L., Zhu, K., Xu, X., Duan, Q., & Zhang, R. (2025). Teg-di: Dynamic incentive model for federated learning based on tripartite evolutionary game. Neurocomputing, 621, 129259.

• Xu, J., Zhou, L., Zhao, Y., Li, X., Zhu, K., Xu, X., Duan, Q., & Zhang, R. (2025). A two-stage federated learning method for personalization via selective collaboration. Computer Communications, 232, 108053.

• Zhu, K., Xu, J., Zhou, L., Li, X., Zhao, Y., Xu, X., & Li, S. (2025). Dmaf: data-model anti-forgetting for federated incremental learning. Cluster Computing, 28(1), 30.