Machine Learning Engineer - Live Content AI (Multimodal)

Responsibilities

Team Introduction:
The Live Streaming Recommendation Core Technology Team is primarily responsible for foundational models and distribution pipelines in live streaming recommendations, including model iteration for modules such as precision ranking, first stage ranking, and recall. The team's key focus is optimizing algorithmic efficiency by addressing business challenges unique to live streaming content, deeply understanding user needs, and building industry-leading recommendation models. Additionally, the team serves as a centralized platform to support iterative improvements across various verticals within live streaming, accelerating business growth.

Responsibilities:
1. Responsible for algorithmic work in Multimodal Learning, Computer Vision (CV), and Natural Language Processing (NLP) for TikTok's global live streaming business. Explore and implement content understanding technologies across various business lines to drive metric growth and technical innovation.
2. Lead the training and optimization of multimodal large models specifically for live streaming visual styles. Leverage capabilities in high-quality data curation and experience in SFT & RL to achieve SOTA performance in business scenarios.
3. Responsible for building and optimizing the fully automated review architecture for content moderation scenarios. Explore innovative applications of large models within the safety and moderation business.
4. Continuously iterate on the understanding and representation capabilities of live content at various granularities. Explore multimodal recommendation technologies and promote their technical deployment in live streaming recommendation scenarios.
5. Build robust content understanding technical capabilities, explore efficient few-shot recognition and model distillation techniques to support diverse business needs in live streaming (e.g., content diversity recognition, interaction understanding, professionalism modeling).

Qualifications

Minimun Qualifications:
1. Master technologies and application experience related to multimodal understanding. Possess a deep understanding of the architecture and training frameworks of mainstream large models, and continuously track the latest advancements in related fields.
2. Passionate about business with a strong sense of ownership and proactivity. Possess excellent communication skills and a strong team spirit.
3. Creative and independent thinker with outstanding skills in experimental analysis and problem-solving. Capable of proposing innovative ideas and validating them through practical implementation.

Preferred Qualifications:
1. At least 2 yrs experience at large scale product oriented IT companies.
2. Candidates who have led key projects or published papers in top-tier conferences/journals in the fields of Multimodal Learning or Large Models are preferred.
3. Experience in the end-to-end deployment of content understanding technologies in Recommendation, Search, or Moderation scenarios is a plus.

Responsibilities

Responsibilities

Team Introduction:
The Live Streaming Recommendation Core Technology Team is primarily responsible for foundational models and distribution pipelines in live streaming recommendations, including model iteration for modules such as precision ranking, first stage ranking, and recall. The team's key focus is optimizing algorithmic efficiency by addressing business challenges unique to live streaming content, deeply understanding user needs, and building industry-leading recommendation models. Additionally, the team serves as a centralized platform to support iterative improvements across various verticals within live streaming, accelerating business growth.

Responsibilities:
1. Responsible for algorithmic work in Multimodal Learning, Computer Vision (CV), and Natural Language Processing (NLP) for TikTok's global live streaming business. Explore and implement content understanding technologies across various business lines to drive metric growth and technical innovation.
2. Lead the training and optimization of multimodal large models specifically for live streaming visual styles. Leverage capabilities in high-quality data curation and experience in SFT & RL to achieve SOTA performance in business scenarios.
3. Responsible for building and optimizing the fully automated review architecture for content moderation scenarios. Explore innovative applications of large models within the safety and moderation business.
4. Continuously iterate on the understanding and representation capabilities of live content at various granularities. Explore multimodal recommendation technologies and promote their technical deployment in live streaming recommendation scenarios.
5. Build robust content understanding technical capabilities, explore efficient few-shot recognition and model distillation techniques to support diverse business needs in live streaming (e.g., content diversity recognition, interaction understanding, professionalism modeling).

Qualifications

Qualifications

Minimun Qualifications:
1. Master technologies and application experience related to multimodal understanding. Possess a deep understanding of the architecture and training frameworks of mainstream large models, and continuously track the latest advancements in related fields.
2. Passionate about business with a strong sense of ownership and proactivity. Possess excellent communication skills and a strong team spirit.
3. Creative and independent thinker with outstanding skills in experimental analysis and problem-solving. Capable of proposing innovative ideas and validating them through practical implementation.

Preferred Qualifications:
1. At least 2 yrs experience at large scale product oriented IT companies.
2. Candidates who have led key projects or published papers in top-tier conferences/journals in the fields of Multimodal Learning or Large Models are preferred.
3. Experience in the end-to-end deployment of content understanding technologies in Recommendation, Search, or Moderation scenarios is a plus.