Job Description

We are looking for talented individuals to join our team. As a graduate, you will get opportunities to pursue bold ideas, tackle complex challenges, and unlock limitless growth.

Successful candidates must be able to commit to an onboarding date by the end of the year. Please state your availability and graduation date clearly in your resume.

Team Introduction: Join the TikTok Engine & Tools AI-3D team, where we are building a next-generation AI-native graphics engine for games and interactive experiences across TikTok and future creative platforms. Our work spans 3D asset and scene generation, character animation, physics simulation, neural rendering, and real-time interactive systems, powered by multimodal AI models. We also develop the data pipelines, runtimes, and evaluation frameworks needed to create persistent, controllable, and responsive digital worlds. By combining expertise in machine learning, computer graphics, and large-scale systems, we transform cutting-edge research into practical technologies for the next generation of interactive content.

Topic Content: Graphics rendering serves as core infrastructure for multimedia applications. Amid the rapid growth of short-video and interactive entertainment, the traditional production pipeline—marked by long cycles, high costs, and steep technical barriers—has severely limited creative expression for mainstream users. In recent years, breakthroughs in AIGC (such as emerging representations like NeRF and 3DGS) have driven 3D generation from experimental research toward commercial deployment, establishing it as a critical foundation for future spatial intelligence.

This topic focuses on cutting-edge 3D generation technologies, aiming to evolve the 3D production pipeline from the traditional "modeling-simulation-rendering" workflow to a new AIGC-driven paradigm: "AI Objects – AI Scenes – AI Animation – AI Rendering." Through this transformation, the project seeks to drastically lower barriers to 3D creation, fully empower the interactive entertainment ecosystem, and build a robust technical foundation for spatial intelligence.

Minimum Qualifications:

- Individuals who are completing or have recently completed a PhD degree in computer science, computer engineering, electrical engineering, applied mathematics, or a related discipline.

- Research experience through a laboratory, thesis, publication, internship, open-source project, or substantial independent project.

- Strong, up-to-date foundation in machine learning, generative modeling, and experimental methodology, with hands-on experience building, training, fine-tuning, or evaluating models using frameworks such as PyTorch or JAX.

- Familiarity with modern generative architectures and training methods such as variational autoencoders, latent tokenizers, diffusion or flow-matching models, diffusion transformers, autoregressive models, multimodal transformers, long-context modeling, pre-training, or post-training.

- Engineering fluency with model development workflows and the ability to reason about model architecture, training dynamics, data quality, and evaluation results.

- Deep experience in at least one of the following AI-for-graphics areas: 3D modeling, asset generation, or 3D representations; animation or motion generation; simulation or physics-aware learning; rendering or neural rendering; world models or video prediction; or graphics and game-engine workflows.

- Ability to understand recent research, design controlled experiments, analyze failures, and contribute to model training, data, or evaluation workflows. Interest in collaborating with Software Engineers to move research prototypes into interactive engine pipelines.

Preferred Qualifications:

- Peer-reviewed publications at leading venues such as SIGGRAPH, CVPR, ICCV, ECCV, NeurIPS, ICML, ICLR, RSS, CoRL, or related research communities.

- Strong technical reports, open-source contributions, demos, project pages, or research projects that show technical depth, originality, and independent experimentation.

- Experience with 3D or world generation conditioned on text, images, video, actions, or structured scene inputs.