Job Description

Our team's mission is to empower content understanding for TikTok Short Video business. We focus on cutting-edge research in content understanding and the development of advanced LLM/MLLM algorithms and applications, including generative recommendation, weakly-supervised learning, few-shot classification, video tagging, multi-task learning, multilingual learning, multimodal pretraining, and more. We aim to succeed both in driving measurable business impact (e.g., recommendation metrics) and delivering state-of-the-art research outputs.

Responsibilities

1. Lead multimodal algorithm development for TikTok’s short-video business, explore applications of multimodal technologies in recommendation systems and other scenarios to improve key business metrics.

2. Conduct cutting-edge research in multimodal and MLLM technologies, design advanced algorithms to solve business requirements while achieving technical breakthroughs.

3. Drive engineering deployment and implementation, ensuring model stability, scalability, and efficiency in production environments.

4. Focus on key areas including (but not limited to):

- General AI platform design and development, including few-shot/zero-shot on MLLM, AI-labeling, auto prompting, active-learning, continue pretraining and RL.

- Integration of content understanding with recommendation systems (e.g., UGC ecosystems, cold start, interest exploration, comment understanding).

- Leveraging multimodal techniques to develop next-generation recommendation systems, such as generative models and end-to-end approaches.