Video Generation Research Engineer

Video Generation Research Engineer

Company: XGEN Labs

Location: Shenzhen, China

Employment Type: Full-time

About XGEN Labs

XGEN Labs is advancing Generative World Simulation , moving AI beyond content generation toward world simulation.

We are building models and systems that enable interactive, dynamic, and continuously evolving visual worlds. We are looking for a Video Generation Research Engineer to advance video generation and interactive visual generation, with a focus on real-time interaction, streaming generation, long-horizon temporal consistency, and controllable outputs.

What You'll Do

• Model Architecture & Development: Design and improve video generation models to support interactive generation, streaming video generation and editing, and long-horizon video synthesis.
• Interactive & Controllable Generation: Research action-conditioned generation, interactive video generation, controllable synthesis, and streaming or autoregressive generation strategies.
• Temporal Consistency & Memory: Develop methods to improve long-term temporal consistency, memory mechanisms, and generation stability over extended sequences.
• Training & Post-Training: Design and optimize training pipelines, including supervised fine-tuning (SFT), reinforcement learning (RL), and knowledge distillation, to improve model capabilities.
• Data & Evaluation: Build and curate training datasets, mine challenging cases, and establish closed-loop training and evaluation workflows to drive continuous model improvement.
• Quality & Efficiency Optimization: Improve visual quality, generation efficiency, controllability, and responsiveness, enabling deployment in real-time and streaming applications.
• Cross-functional Collaboration: Work closely with research, engineering, and product teams to bring video generation capabilities into practical products and interactive experiences.

What We're Looking For

• Strong knowledge of modern visual generation models and architectures, particularly diffusion-based models and Diffusion Transformers (DiT).
• Hands-on experience with video generation, diffusion models, or temporal modeling.
• Proficiency in PyTorch and experience implementing, training, or fine-tuning generative models.
• A solid understanding of video generation pipelines, model training strategies, and evaluation methods.
• Strong problem-solving skills and the ability to translate research ideas into working implementations and measurable improvements.

Nice to Have

• Experience with interactive video generation, streaming generation, action-conditioned generation, or controllable video synthesis.
• Experience improving long-horizon temporal consistency, persistent memory, or autoregressive video generation.
• Familiarity with recent video generation models and frameworks, such as Wan, LTX, or other state-of-the-art open-source or proprietary models.
• Experience with post-training techniques, including SFT, reinforcement learning, and distillation.
• Experience building data pipelines, mining hard cases, and developing systematic model evaluation workflows.
• Research publications, open-source contributions, or demonstrable project experience in video generation or interactive visual generation.

Why Join XGEN Labs?

• Work on challenging problems at the intersection of video generation, interactive AI, and world simulation.
• Help advance models beyond static content generation toward dynamic, responsive, and persistent visual experiences.
• Collaborate with research, engineering, and product teams to turn emerging model capabilities into real-world applications.

Interested? Apply with your CV and, where available, links to relevant research papers, GitHub repositories, model demos, or technical projects.