AI LLM Post-Training Engineer Intern

TikTok; San Jose (CA); On-site; USD 51,000 - 73,000; Part time

Benefits offered by this job

Health insurance

Paid holidays

Paid sick time

Housing allowance

Job summary

TikTok in San Jose offers an AI internship focused on post-training methods for LLMs and multimodal AI. You will work with the Multimedia AI team to explore instruction tuning, evaluation pipelines, and alignment strategies while collaborating with engineers to bring research into production-ready applications.

The role suits students pursuing CS/ML degrees, with strong Python skills and experience in PyTorch or JAX. A dynamic, learning-oriented environment awaits.

Qualifications

• Pursuing an Undergraduate or Master’s degree in CS, ML, or related field.
• Strong Python programming and DL framework experience (PyTorch/JAX).
• Foundational understanding of Transformer architectures and LLM training.
• Ability to learn quickly in a fast-paced environment and interest in AI deployment/applications.
• Experience with LLM fine-tuning, RLHF, or synthetic data.
• Familiarity with distributed training tools (DeepSpeed, Megatron-LM).
• Interest in multimodal AI (text with video/audio).
• Record of AI project work or internships.

Responsibilities

• Support post-training strategies (instruction tuning, SFT/DPO/PPO, model alignment).
• Assist in building evaluation pipelines for model performance, helpfulness and safety across multimedia use cases.
• Research and implement reward modeling and human preference learning techniques.
• Collaborate with engineering to bridge research and production-ready AI apps (video understanding, translation, classification).
• Analyze large-scale datasets to identify patterns improving model behavior and alignment.

Skills

Python

PyTorch

Transformer architectures

LLM training principles

Adaptability

Education

Undergraduate or Master’s in CS/ML

Tools

DeepSpeed

Megatron-LM

Distributed training

Job description

TikTok in San Jose offers an AI internship focused on post-training methods for LLMs and multimodal AI. You will work with the Multimedia AI team to explore instruction tuning, evaluation pipelines, and alignment strategies while collaborating with engineers to bring research into production-ready applications.

The role suits students pursuing CS/ML degrees, with strong Python skills and experience in PyTorch or JAX. A dynamic, learning-oriented environment awaits.