Machine Learning Expert, Risk Control - Global E-Commerce

The E-Commerce Risk Control (ECRC) team is missioned:

• To protect our E-Commerce users, including and beyond buyer, seller, creator;
• By securing the integrity of our ecommerce ecosystem and providing a safe shopping experience on the platform;
• Through building infrastructures, platforms and technologies, as well as collaborating with many cross-functional teams and stakeholders.

The ECRC team works to minimize the damage of inauthentic behaviors on our E-Commerce platforms, covering multiple classical and novel community and business risk areas such as account integrity, incentive abuse, malicious activities, brushing, click-farm, information leakage etc.

In this team you'll have a unique opportunity to have first-hand exposure to the strategy of the company in key security initiatives, especially in building scalable and robust, intelligent and privacy-safe, secure and product-friendly systems and solutions. Our challenges are not some regular day-to-day technical puzzles -- You'll be part of a team that's developing novel solutions to first-seen challenges of a non-stop evolvement of a phenomenal product eco-system. The work needs to be fast, transferrable, while still down to the ground to making quick and solid differences.

Responsibilities

1. Advance the foundational capabilities of large language models in understanding structured data and multimodal data, including graphs and text. Explore foundation models tailored to the risk management domain and develop Agentic solutions to automate and enhance risk control workflows. Drive the application of advanced LLM reasoning capabilities across merchant operations, user risk management, platform safety, and other business scenarios.
2. Lead the development of risk-focused foundation models for platforms including Douyin E-commerce, TikTok Shop, and Tokopedia. Address the limitations of current SOTA models in understanding highly adversarial risk content and detecting AIGC-generated content.
3. Develop autonomous, Agent-based adversarial solutions to improve risk detection efficiency and reduce operational costs. Tackle challenges such as the large volumes of data and complex decision rules involved in risk assessment, while developing solutions for cross-modal long-context information extraction, instruction following under complex compliance requirements, runtime model auditing, and autonomous model evolution.
4. Stay current with the latest advances in the field and explore the integration of LLMs with traditional machine learning techniques for practical applications in risk management.

Qualifications Minimum Qualifications

1. Master's degree or above with at least 5 years of relevant industry experience.
2. Strong foundations in machine learning, with an in-depth understanding of content understanding and generation technologies, including deep learning, large language models, multimodal models, and generative models. Solid mathematical skills, strong self-learning ability, and hands-on experience applying these technologies to real-world risk management scenarios are required.
3. Strong programming skills and proficiency in relevant machine learning and engineering frameworks.
4. Experience with multimodal large language models is required. Hands-on experience deploying Agentic systems in production environments is required.

Preferred Qualifications

1. Experience in developing risk control algorithms for internet platforms and managing mature technical teams is a strong plus.
2. Publications at leading computer science conferences or journals—such as NeurIPS, ICML, CVPR, ICCV, ECCV, IJCAI, AAAI, KDD, SIGIR, WWW, ACL, IEEE TPAMI, or IJCV—or strong performance in relevant technical competitions are preferred.