Job Description

About the Team

TikTok AI Search Infra team is responsible for building highly performant, scalable and stable infrastructures that serve billions of search requests everyday, and reinventing the search experience of billions of users globally. We apply cutting edge ML/NLP/LLM/VLM technology for end-to-end modeling, and focus on building and optimizing performance/efficiency for large-scale ML infra and online/offline distributed systems, enabling AI to realize its potential value for billions of TikTok users. We embrace a culture of self-direction, intellectual curiosity, openness, and problem-solving.

Responsibilities

- Evolve TikTok’s AI Search multi-agent LLM engine, supporting ReAct + Tool calling, DAG-based workflows, and RAG paradigms

- Deploy and optimize text/multimodal LLMs, including inference acceleration, model alignment during training, and reinforcement learning

- Build scalable and reliable end-to-end serving platforms supporting multiple TikTok AI Search use cases, such as Q&A cards, in-app chatbot, and visual search, including both online and offline scenarios

- Build large-scale data architecture for handling billion-level data records: offline computation, distributed system performance and scheduling optimization, as well as building high-availability, high-throughput, and low-latency online services

- Design and build personalized AI search capabilities to achieve more accurate AI Overview and Q&A experience for users

- Collaborate with modeling and product teams to deliver better AI search experience for TikTok users