PhD Intern, Vision Engineering – GenAI Optimization

Job Description

Efficient ML Engineer Research Intern (AI Platform) – 2026 Start (PhD) | TikTok

The Tone:
This is a PhD internship for a 2026 start at TikTok, located in Los Angeles, CA. The Vision Engineering Team at TikTok delivers GenAI technologies directly into TikTok products worldwide, leveraging proprietary AI infrastructures to streamline creation, integration, testing, and deployment of these features. This role is crucial for enhancing user experience by powering diverse functionalities like visual enhancements, video editing tools, and creative camera filters. The 12-week internship offers students hands-on experience, the opportunity to develop fundamental skills, explore potential career paths, and actively contribute to the company’s products, research, and future technologies.

The TL;DR
• Role: Internship
• Location: In-person, Los Angeles, CA
• Pay: $60 hourly
• Team: Vision Engineering Team
• Mission: Optimize GenAI models and AI infrastructure for efficiency and performance to enhance TikTok’s product capabilities.
• Tech Stack: C++, Python, Pytorch, DeepSpeed, Jax

What You’ll Actually Do
• Development: Develop algorithm acceleration technologies for text-to-image/text-to-video models using knowledge distillation, model architecture redesign, and parameter-efficient design for significant efficiency gains.
• Innovation: Lead generative model innovation with a focus on diffusion acceleration through sampling step reduction and latent optimization, as well as autoregression model efficiency.
• Collaboration: Collaborate with other teams to identify performance bottlenecks within existing vision models.
• Optimization: Optimize vision models through algorithmic breakthroughs to improve performance.
• Enhancement: Enhance ByteDance’s product capabilities by integrating optimized vision models and GenAI features.

The Must-Haves
• Background: Currently pursuing a PhD in Computer Science, engineering, or a quantitative field.
• Experience: Expertise in diffusion models (Stable Diffusion/DiT) with a deep understanding of computational bottlenecks and optimization methodologies; proven experience in at least one of the following: model compression (quantization/knowledge distillation), efficient architectures (MoE/sparse attention), or generative alignment (RLHF/DPO).
• Skills: Proficient in C++/Python and high-performance coding; excellent communication and teamwork skills.
• Bonus: Kaggle competition achievements, publications at ICML/NeurIPS/CVPR, or open-source contributions (e.g., HuggingFace Diffusers optimization); research experience in GenAI /MLsys areas; familiarity with open-source deep learning frameworks such as Pytorch/DeepSpeed/Jax.