LLM & Agent Algorithm Graduate – Search (2027 Start)

About the role

Join TikTok's TT-Search Algorithm & Applied AI team, the core engine behind TikTok Search, to build the next generation of search powered by large language models and autonomous agents. As a PhD graduate you will work on cutting‑edge research and engineering to shape the future of search and multimodal AI.

Key responsibilities

• Develop and optimise the search‑domain LLM foundation, including ultra‑long‑text pre‑training and multimodal representation learning.
• Build the Agent execution framework (Harness) for tool calling, planning, memory, and environment interaction, and design multi‑agent orchestration algorithms.
• Research long‑term memory mechanisms, self‑evolution loops, and recursive self‑improvement pipelines for LLMs.
• Create online reward/verifier systems and automated evaluation frameworks for complex, long‑horizon tasks.
• Implement business‑focused agents for multimodal AIGC creation, visual search, on‑device intelligence, and other search‑related products.

Required profile

• PhD in Computer Science or a related field, completed or near completion.
• Strong foundation in machine learning/deep learning with familiarity of state‑of‑the‑art LLM and agent technologies.
• Experience with PyTorch or TensorFlow, distributed and mixed‑precision training, and model compression techniques.
• Track record of publications at top conferences (NeurIPS, ICML, ICLR, CVPR, ACL, etc.) or notable competition awards.

Required skills

• Machine Learning & Deep Learning
• Large Language Models (LLM) and Agent systems
• PyTorch / TensorFlow
• Distributed training, mixed‑precision training
• Model quantisation, pruning, distillation, TensorRT
• Big‑data frameworks (MapReduce, Spark)

What we offer

• Positive team atmosphere
• Access to industry experts
• Meals provided