The opportunity to work on TikTok’s audio and video creation experiences and build the next generation of AI-powered creative tools for users worldwide.
A role centered on real product challenges and end-to-end delivery, without being siloed into Android- or iOS-only development.
The opportunity to integrate AI coding deeply into everyday engineering and help shape the next generation of software development practices....
Participate in building large-scale (10 million to 100 million) e-commerce recommendation algorithms and systems, including commodity recommendations, live stream recommendations, short video recommendations etc in TikTok.
Build long and short term user interest models, analyze and extract relevant information from large amounts of various data and design algorithms to explore users' latent interests efficiently.
Design, develop, evaluate and iterate on predictive models for candidate generation and ranking(eg. Click Through Rate and Conversion Rate prediction) , including, but not limited to building real-time data pipelines, feature engineering, model optimization and innovation....
Education & Foundation: Ph.D. in Computer Science, AI, Mathematics, or a related field, with a strong foundation in data structures, algorithms, and mathematical modeling.
AI/ML Expertise: Solid understanding and research experience in Deep Learning, NLP, CV, Reinforcement Learning, Generative Models, or Multimodal Learning.
Coding & Engineering: Proficient in major programming languages and machine learning frameworks (e.g., PyTorch, TensorFlow), combined with excellent problem-solving, self-learning, and teamwork skills....
Build and evolve the agent runtime (harness / agent loop) powering our creator / seller agents — orchestrate skills, tools, and context.
Engineer context & memory for long multi-turn agents — agentic memory (structured note-taking), context compaction / summarization, context editing / observation masking, and just-in-time (retrieve-then-load) retrieval.
Post-train and fine-tune LLMs (SFT / DPO / RL) and build the data flywheel that turns served conversations into training / eval / retrieval signals....
Recommendation Large Models: Address challenges such as gradient convergence and representation drift in ultra-long behavioral sequences, enabling the system to achieve true ""logical reasoning"" capabilities.
Unified Multimodal Semantic Space: Explore alignment across video, image-text content, and user intent, constructing a fully multimodal semantic space that goes beyond text.
Agentic Rec:Develop recommendation agents with capabilities such as self-reflection, tool invocation, and long-horizon planning, driving a transformation of recommendation and content distribution experiences....
