Machine Learning Engineer Intern (Risk & Integrity) - 2027 Start (PhD)

About the team

The USDS Platform and Community Integrity (PaCI) team is the architectural backbone of TikTok's Risk & Integrity in the US. We drive end-to-end integrity for TikTok and its expanding ecosystem, including Lemon8 and emerging affiliate platforms, by leveraging localized US data and state-of-the-art AI/LLM capabilities.

We are looking for talented individuals to join us for an internship. Our internship program offers students hands-on experience, industry exposure, and opportunities to apply their knowledge to real-world challenges while building a strong foundation for personal and professional growth.

Interns will gain practical experience, explore potential career paths, and participate in social events, learning programs, and development workshops alongside industry professionals.

Applications are reviewed on a rolling basis, so we encourage you to apply early. Please clearly state your availability in your resume, including your start and end dates.

Responsibilities

• Work with large-scale data, logs, metrics, or experiments to understand user search behavior and improve product or system quality.
• Collaborate with engineering, product, data science, and machine learning partners to define requirements and deliver project milestones.
• Write clear, maintainable code and documentation for assigned workstreams.
• Communicate progress, risks, and learnings with mentors and stakeholders throughout the internship.

Minimum Qualification(s)

• Currently pursuing a PhD degree in Computer Science, Computer Engineering, Software Engineering, or a related technical field.
• Able to commit to a 12-week internship during Summer 2027.
• Solid foundation in Machine Learning; proficient in applying Transformer and GNN architectures to solve large-scale integrity challenges with social graph and behavior sequence data. Experience designing model evaluation systems that balance technical performance with product goals.
• Strong engineering and coding skills. Proficient in at least one language among Python, C++, or Go; familiar with distributed computing (e.g., Spark, Flink) and deep learning frameworks (e.g., PyTorch).

Preferred Qualifications

• Agentic AI Experience: Hands-on experience building LLM-based Agents capable of task planning, tool-use, or closed-loop autonomous decision-making.
• LLM Security & Alignment: Experience in LLM Post-training or Fine-tuning (e.g., SFT, RLVR, RLHF) to optimize models for specific risk domains like fraud intent recognition or malicious redirection.
• Multimodal Expertise: Proven track record in video understanding or multimodal representation; ability to integrate visual, audio, and textual features to solve complex behavioral modeling challenges.