Job Description

About the team

Our Search Team is responsible for building and owning TikTok's search engine which provides our users the best search experience. On the TikTok Search Team, you’ll have the opportunity to build a full-stack search engine system and combine information retrieval technology with modern machine learning methods from related fields such as NLP, Computer Vision, Multimodal, and Recommender Systems. We embrace a culture of self-direction, intellectual curiosity, openness, and problem-solving.

Job Responsibilities

Be responsible for the iterative optimization of photo-text search quality and relevance algorithm directions. Based on multi-modal technology for photo-text content understanding, grasp the core characteristics of the photo-text genre, focus on optimizing photo-text relevance models, solve core pain points such as photo-text semantic alignment and cross-modal matching, break through technical bottlenecks, expand the contribution of photo-text search to the search business, and create a high-quality photo-text search experience that meets user needs.

Core Responsibilities:

1. Photo-text multi-modal understanding: focus on solving core pain points such as photo-text semantic alignment and cross-modal matching deviation, and integrate photo-text visual features, text semantics and other data.

2. Be responsible for the optimization of photo-text relevance and quality, and promote the experience improvement of photo-text in search distribution by combining LLM/VML technologies.

3. Engineering implementation and algorithm performance optimization: be responsible for the engineering implementation of photo-text search algorithm models, complete the full-process optimization of model training, inference and deployment, and ensure the online stability of algorithms.

4. Track cutting-edge technology trends, carry out technical pre-research and innovative practices combined with photo-text search business scenarios, explore in-depth applications in photo-text semantic understanding and relevance matching, form technical precipitation, and improve the technical competitiveness of the team.