Position Overview We are seeking an experienced AI Research Scientist to lead foundation model development initiatives. The ideal candidate will have hands-on experience in training large-scale models at major tech companies and a proven track record in advancing the state-of-the-art in foundation models. Key Responsibilities Lead the architecture design and training of large-scale foundation models Develop and optimize model training pipelines for distributed systems Drive research initiatives in model scaling, efficiency, and performance Implement innovative approaches to improve model capabilities and training efficiency Collaborate with the engineering team to productionize research breakthroughs Guide technical decisions related to model architecture and training strategies Required Qualifications Ph.D. in Computer Science, Machine Learning, or related field 3+ years of experience in training large-scale models at major tech companies, including: International tech leaders (e.g., Google, Meta, Microsoft, OpenAI, Anthropic) OR Leading Chinese tech companies (e.g., ByteDance, Alibaba, Baidu, Tencent, SenseTime, Huawei) Proven experience with distributed training systems and large-scale model optimization Deep understanding of transformer architectures and their variants Strong track record in developing and training foundation models Extensive experience with PyTorch and/or JAX Publication record in top-tier conferences (NeurIPS, ICML, ICLR) Preferred Qualifications Experience with both Chinese and international AI ecosystems Familiarity with Chinese AI infrastructure (e.g., ModelArts, PAI, ByteMLab) Background in scaling laws and efficient training strategies Experience with video generation models or multimodal architectures Track record of open-source contributions to major ML frameworks Experience with ML infrastructure design and implementation Familiarity with mixed-precision training and model parallelism Experience with custom CUDA kernels and optimization Technical Expertise Large-Scale Training: Distributed training frameworks, model parallelism strategies Infrastructure: International cloud platforms (AWS/GCP) Chinese cloud platforms (Alibaba Cloud, Tencent Cloud, Huawei Cloud) Languages: Python, CUDA, C++ (optional) Frameworks: Standard: PyTorch, JAX, DeepSpeed, Megatron-LM Chinese ecosystem: PaddlePaddle, MindSpore (plus) Development Tools: Git, Docker, Kubernetes Monitoring: Weights & Biases, MLflow, or similar tools What We Offer Opportunity to shape the future of foundation models in video generation Leadership role in technical decision-making Access to substantial computing resources and infrastructure Competitive compensation package including equity Regular collaboration with top researchers in the field Support for conference attendance and research publication International exposure and collaboration opportunities Location Hong Kong (on-site, Hong Kong Science and Technology Park) Expected Impact Drive the development of next-generation foundation models Lead research initiatives that push the boundaries of model capabilities Build and mentor a world-class research team