AI Research Scientist

AI Research Scientist - Video Generation

About Us

We are a cutting-edge AI startup specializing in next-generation video generation technology based in Hong Kong Science and Technology Park. Our mission is to push the boundaries of what's possible in AI-driven video generation through innovation of foundation model. As a growing startup located at Hong Kong Science Park, we offer a dynamic research environment where your contributions can advance the state-of-the-art in AI technology.

Position Overview

We are seeking a dedicated AI Research Scientist to join our foundation model research team. The ideal candidate will have hands-on experience in training large-scale models and a strong passion for advancing fundamental research in foundation models. This position is designed to support early-career researchers in developing their expertise while contributing to cutting-edge AI research.

Key Research Responsibilities

• Conduct fundamental research in large-scale foundation model architectures and training methodologies
• Design and implement novel approaches to video generation using transformer-based architectures
• Develop and optimize model training algorithms for distributed computing environments
• Investigate scaling laws and efficiency improvements for foundation model training
• Research innovative techniques to enhance model capabilities and performance
• Implement and evaluate state-of-the-art methods in multimodal AI and video synthesis
• Collaborate with research team members on joint research projects and publications
• Contribute to the development of proprietary research datasets and evaluation metrics
• Conduct experimental validation of theoretical approaches and document findings
• Participate in research discussions and present findings at internal research meetings
• Publish research paper on top-tier AI/ML conferences or journals

Required Qualifications

• Ph.D. in Computer Science, Machine Learning, Artificial Intelligence, or related field
• 2+ years of research experience in large-scale model training, preferably at:
• International research institutions or tech companies (e.g., Google Research, Meta AI, Microsoft Research, OpenAI, Anthropic) OR
• Leading Chinese research labs or tech companies (e.g., ByteDance AI Lab, Alibaba DAMO Academy, Baidu Research, Tencent AI Lab, SenseTime Research, Huawei Noah's Ark Lab)
• Strong research background in distributed training systems and large-scale model optimization
• Deep understanding of transformer architectures, attention mechanisms, and their variants
• Proven research experience in developing and training foundation models
• Proficiency in PyTorch and/or JAX for research implementation
• Publication record in top-tier AI/ML conferences (NeurIPS, ICML, ICLR, CVPR, ICCV) or strong research portfolio

Preferred Research Qualifications

• Research experience bridging Chinese and international AI research communities
• Knowledge of Chinese AI research infrastructure and platforms (e.g., ModelArts, PAI, ByteMLab)
• Research background in scaling laws and efficient training methodologies
• Specialized research experience in video generation models or multimodal architectures
• Active contributions to open-source research projects and ML frameworks
• Research experience in ML systems and infrastructure optimization
• Knowledge of mixed-precision training techniques and model parallelism strategies
• Experience with custom CUDA kernel development for research applications

Technical Research Skills

• Foundation Model Research: Transformer architectures, attention mechanisms, scaling methodologies
• Distributed Training Research: Multi-GPU/multi-node training, model parallelism, gradient synchronization
• Cloud Research Infrastructure:
• International platforms (AWS/GCP for research computing)
• Chinese platforms (Alibaba Cloud, Tencent Cloud, Huawei Cloud)
• Programming: Python, CUDA, C++ (beneficial for optimization research)
• Research Frameworks:
• Standard: PyTorch, JAX, Hugging Face Transformers, DeepSpeed, Megatron-LM
• Chinese ecosystem: PaddlePaddle, MindSpore (advantageous)
• Research Tools: Git, Docker, Jupyter, Weights & Biases, MLflow, TensorBoard

Research Environment & Support

• Access to substantial GPU computing resources for large-scale experiments
• Opportunity to conduct cutting-edge research in video generation and foundation models
• Collaborative research environment with opportunities for co-authored publications
• Support for attending international AI/ML conferences and workshops
• Access to research datasets and computational infrastructure
• Mentorship opportunities with senior researchers in the field
• Regular research seminars and technical discussions
• International collaboration opportunities with leading research institutions

Research Impact Goals

• Advance the state-of-the-art in foundation models for video generation
• Contribute to high-impact research publications in top-tier venues
• Develop novel methodologies that can be translated into practical applications
• Build expertise in large-scale AI model research within the Hong Kong innovation ecosystem

Application Requirement

Please submit:

1. Comprehensive CV highlighting research experience and publications
2. Research statement (2-3 pages) outlining your research interests and vision for foundation model research
3. List of significant research projects with detailed descriptions of your contributions
4. Links to Google Scholar profile, GitHub repositories, and any published code
5. Contact information for 2-3 research references

To apply or inquire about this research position, please contact [Confidential Information].