大模型Agent算法工程师-TikTok搜索（北京/上海）
Core
Develop and iterate the foundational Large Language Model (LLM) and Agent infrastructure for TikTok Search, enabling precise search, multi-modal understanding, and autonomous agent execution.
Role type
Senior IC Machine Learning Engineer (LLM & Agent Systems)
Builds
Search LLM base, Agent execution framework (Harness), long-memory mechanisms, and automated evaluation systems for TikTok Search.
Domain
Search, Large Language Models, Multi-modal AI, Agent Systems
Deliverable
production ML models
Required skills
LLM pre-training and fine-tuning, multi-modal representation learning, agent planning and tool use, multi-agent orchestration, inference optimization (quantization, pruning, distillation), distributed training, automated data synthesis and evaluation
Preferred skills
Next-gen non-Transformer architecture research, recursive self-improvement (RSI) mechanisms, visual search and OCR, AGI safety and alignment
Technologies
Transformer architectures, TensorRT, MR/Spark, RLHF, SFT, RSI
Responsibilities
Lead R&D of search LLM base and pre/post-training pipelines; build Agent Foundation platform for multi-agent scheduling and collaboration; design long-memory and self-improvement loops; construct data synthesis and automated evaluation systems; drive business deployment of long-term task agents and visual search capabilities.
Seniority
Senior, hands-on IC