I am currently a second-year Master’s student in the School of Computer Science and Engineering at Southeast University, specializing in Human-AI Interaction and Multi-Agent Systems. I am supervised by Associate Prof. Zhuying Li at the PALM Lab.

Education

• MSc in Computer Science and Engineering, Southeast University, China
  2024 - 2027 (expected)

• BSc in Software Engineering, Southeast University, China
  2020 - 2024

• Exchange Program in Computer Science, KTH Royal Institute of Technology, Sweden
  Jan. 2023 - Jun. 2023

Research Interests

• Human-AI Collaboration and AI Scaffolding
• Human-Computer Interaction (HCI)
• Multi-Agent Systems
• Generative AI Systems and Efficient Inference

Working Experience

• TikTok Engineer Intern, ByteDance, Beijing, China
  Apr. 2026 - Present
  • Working on finetuning SEED-1.8 model for creative asset selection for user growth.
  • Developed and shipped the ad budget allocation system, enabling automated budget distribution across ad campaigns and countries.
  • Building a creative asset manager agent for user growth scenarios, integrating LLM reasoning with tool-based workflows for asset retrieval, analysis, planning, and management.
• AI Frameworks Engineer Intern, Intel Corporation, Shanghai, China
  Nov. 2025 - Apr. 2026
  • Adapted and optimized 15+ multimodal AIGC models and end-to-end generation pipelines, covering text-to-image, image editing, text/image-to-video, speech synthesis and cloning, 3D generation, and video super-resolution.
  • Contributed to SGLang Diffusion support for Intel XPU, including XCCL distributed communication adaptation and Triton kernel fallback paths for diffusion model inference.
  • Built DiT inference acceleration for ComfyUI with block-level output caching and reuse, achieving 1.3x-2.0x speedup while preserving generation quality, and up to 2.2x when combined with torch.compile.
  • Supported GGUF quantized inference on Intel XPU through SYCL custom operators, reducing memory pressure for large generative models.
  • Implemented Raylight-based multi-GPU video generation for Wan2.2 14B and HunyuanVideo 1.5.
• Co-Founder & AI Platform Engineer, INFERA, Shenzhen, China
  Mar. 2025 - Aug. 2025
  • Participated in the design and development of InfNest, a proactive AI and multi-agent platform for next-generation edge hardware and AI glasses.
  • Developed the InfNest multi-agent framework with Master/Member/Group hierarchy, a three-layer memory system, MCP-based tool extension, and a dynamic workflow engine for complex task orchestration.
  • Contributed to backend development for the Melons AI headphone project, implementing AI music services.
• AI Frameworks Engineer Intern, Intel Corporation, Shanghai, China
  Sep. 2023 - May. 2024
  • Contributed to the development and optimization of ipex-llm, an open-source project for improving large language model fine-tuning and inference on Intel GPU, NPU, and CPU.
  • Implemented and integrated efficient parameter fine-tuning methods, including QLoRA, LISA, and GaLoRA, for mainstream models such as LLaMA and Mistral.
  • Supported inference adaptation for mainstream LLMs, including low-bit quantization, Mixtral-8x7B sparse MoE inference optimization, and speculative decoding to improve inference throughput.
  • Supported enterprise deployment and performance tuning, including Docker- and Kubernetes-based distributed inference and fine-tuning environments.
• C++ Developer Intern, China Telecom Corporation Limited, Nanjing, China
  Jul. 2023 - Aug. 2023
  • Worked on streaming media applications and system concurrency optimization.

Selected Projects

• ComfyUI-CacheDiT, 280+ GitHub stars
  Jan. 2026 - Feb. 2026
  • Designed a warmup-skip caching strategy for Diffusion Transformer denoising, reusing block-level outputs across adjacent steps to accelerate image and video generation while preserving visual quality.
  • Implemented model-specific caching strategies for LTX-2 dual-latent video/audio generation and Wan2.2 MoE high/low-noise expert pipelines.
  • Built automatic model detection and zero-configuration ComfyUI nodes for mainstream DiT models.
• Optical Flow and YOLO-based Underwater Fish Detection, National College Student Innovation and Entrepreneurship Project
  Dec. 2021 - May. 2023
  • Proposed an optical-flow-YOLO spatiotemporal fusion detection algorithm using Gaussian Mixture Models and optical flow to extract motion features in complex underwater scenes.
  • Built an end-to-end real-time fish detection and tracking framework, achieving 25+ FPS and improving robustness under low-light and turbid-water conditions.

Awards and Honors

• First-class Postgraduate Scholarship, Southeast University (2024)
• President’s Scholarship, Southeast University (2021)
• Huawei Scholarship (2023)
• Zhishan Scholarship, Southeast University (2022)
• Excellent Student Leader, Southeast University (2022)
• Merit Student, Southeast University (2021)

Publications

• Mingtao Wu, Ziteng Zhang, Mengjie Tang, Zhuying Li, and Yan Wang. GreenCompass: Weaving Playful Nature Engagement into Urban Micro-Moments through Context-Aware Gamification. Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI ‘26). Paper
• Zhuying Li, Ziteng Zhang, Xin Sun, and Yan Wang. 2026. Digital Spirituality in Mainland China: Understanding Online Practices for Designing Culturally Relevant Spiritual Experiences. Proc. ACM Hum.-Comput. Interact. 10, 2, Article CSCW013 (CSCW ‘26). Paper
• Mingtao Wu, Zhuying Li, Ziteng Zhang, and Yan Wang. 2025. GreenCompass: Experiencing Urban Pocket Time as Playful Nature Connection through Context-Aware Gamification. In Proceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems (CHI EA ‘25). Paper