LLM Post-Training Engineer — Graduate

TikTok; San Jose (CA); Hybrid; USD 128,000 - 317,000; Full time

Job summary

TikTok in San Jose is seeking a graduate-level candidate to help advance post-training strategies for large language models, including instruction tuning and alignment, while helping bridge research with production.

Work with the Multimedia AI team on evaluating model performance, safety, and multilingual/multimodal tasks such as video understanding and translation in a fast-paced environment.

Qualifications

• Master's degree in Computer Science, ML, or related field.
• Proficiency in Python and DL frameworks (PyTorch, JAX).
• Strong understanding of Transformer models and LLM training.
• Ability to learn quickly in a fast-paced AI environment.

Responsibilities

• Support post-training strategies (instruction tuning, SFT/DPO/PPO) and alignment.
• Build evaluation pipelines to measure model performance, helpfulness, safety.
• Research and implement reward modeling and human preference learning.
• Collaborate with engineering to bridge research and production (video understanding, translation, content classification).
• Analyze large-scale datasets to identify patterns improving model behavior.

Skills

Python

PyTorch

JAX

Transformer architectures

LLM training principles

Fast learner

Education

Master's degree in Computer Science / Machine Learning

Tools

DeepSpeed

Megatron-LM

Job description

TikTok in San Jose is seeking a graduate-level candidate to help advance post-training strategies for large language models, including instruction tuning and alignment, while helping bridge research with production.

Work with the Multimedia AI team on evaluating model performance, safety, and multilingual/multimodal tasks such as video understanding and translation in a fast-paced environment.