Interactive Vision Model Research Intern

About the role

Lightspeed Studios is a Tencent business unit advancing game development through stories, gameplay and advanced technology. Its Lightspeed Tech Center is an R&D department under Lightspeed Studios that develops PUBG Mobile and other games, and leads research and technical services across engine, audio, QA, AI and next-generation game technologies. This is an interactive vision model research internship within the center.

Key responsibilities

• Design and iterate on the next-generation video generation foundation architecture.
• Explore core technologies for long video generation, including long-context attention, KV Cache compression and memory mechanisms.
• Research high-compression-ratio video tokenizers and advance unified modeling for multi-resolution and multi-frame-rate videos.
• Lead video pre-training at large token scale, defining data mixture and curriculum learning strategies.
• Explore scaling laws, define scientific scale-up paths and improve key model capabilities.
• Track industry state-of-the-art, run comparative experiments, drive technical roadmap decisions and publish academic papers.

Required profile

• Currently pursuing a Ph.D. in AI-related fields with first-author papers at top-tier conferences.
• Experience training video/image generation models from scratch; 3+ years of relevant research experience.
• Deep understanding of the design trade-offs in Video VAEs and tokenizers.
• Publications related to video generation or diffusion models.
• Experience in core industry product R&D or hands-on work with the latest technologies is preferred; experience leading end-to-end video foundation model projects is preferred.

Required skills

• Proficiency in the principles and engineering implementation of diffusion models and autoregressive generation.
• Highly proficient in PyTorch and large-scale distributed training.