Research Scientist Intern, Scam and Trust (PhD)

Meta was created to foster connection and sharing worldwide. Over the past decade, our platforms have transformed global communication. With over a billion active users, joining Meta provides extraordinary opportunities to make a meaningful impact within a rapidly evolving organization.

Meta is seeking Research Scientist Interns to join our Scam and Trust team within Facebook. Scams are a global financial threat. At Meta’s scale, proactive detection and prevention are essential for protecting billions of users worldwide, making addressing scams one of the top priorities at Meta. The space is highly adversarial as bad actors continually change tactics and operate across content, behavior, and social networks, making isolated signals and static models insufficient. You will develop novel modeling and evaluation methods that address these challenges.

You will work alongside research scientists and engineers to translate a real-world scam-detection and prevention challenge into a focused research project. Research opportunities include combining large language model (LLM) reasoning with actor representations and content or behavioral signals, and developing agentic systems that use tools to gather evidence, investigate suspicious activity, and adapt their reasoning.

You may also investigate recursive self-improvement methods that enable LLM-based systems to identify weaknesses, propose improvements to their strategies or workflows, and validate those improvements through rigorous evaluation. Other research opportunities include learning from noisy and limited labels, transferring methods across surfaces, and maintaining robustness under distribution shift as adversaries evolve. Promising research may inform production systems that protect people from scams and contribute to research publications.

Our internships are twelve (12) to sixteen (16), or twenty-four (24) weeks long and we have various start dates throughout the year.

Responsibilities

Formulate and investigate a focused research question in adversarial scam detection and prevention.
Develop LLM reasoning and representation-learning methods that combine content, behavioral, or social-network evidence to identify scams and distinguish malicious activity from legitimate behavior.
Research agentic systems that plan investigations, use tools, and synthesize evidence, evaluating both individual decisions and end-to-end effectiveness.
Investigate recursive self-improvement and feedback-driven methods that improve agent strategies, reasoning, or workflows over successive iterations, with controlled evaluation to distinguish genuine gains from overfitting or error amplification.
Design rigorous experiments to evaluate robustness under noisy labels and distribution shift, detection–false-positive trade-offs, and computational efficiency.
Collaborate with research scientists, engineers, and cross-functional partners to prototype promising approaches, assess their applicability to production, and communicate research plans, progress, and results.
Document research findings and, where appropriate, contribute to publications and research that can inform Meta product development.

Qualifications

Currently has or is in the process of obtaining a Ph.D. degree in Computer Science, Artificial Intelligence, Natural Language Processing, Speech Recognition, Sentiment Analysis, Computer Vision, or relevant technical field
Must obtain work authorization in country of employment at the time of hire and maintain ongoing work authorization during employment
Experience with deep learning frameworks
Experience building systems based on machine learning, deep learning methods, or natural language processing
Familiarity with algorithms behind generative model training and common deep learning architectures (Transformers, BERT etc.) Intent to return to the degree program after the completion of the internship/co-op
Proven track record of solid research achievements as demonstrated by grants, fellowships, patents, as well as publications at leading AI conferences such as NeurIPS, ICML, ICLR, ACL, EMNLP and KDD
Research or project experience in one or more of the following: LLM reasoning, agentic systems and tool use, recursive self-improvement, reinforcement learning, automated evaluation, learning from feedback, behavioral or graph representations, weak supervision, or transfer learning. Experience applying these methods to adversarial or other real-world problems is a plus
Demonstrated software development experience via tech internships, work experience, coding competitions, or widely used contributions in open source machine-learning repositories
Experience working and communicating cross functionally in a fast-paced team environment. Ideal candidates should have the ability to quickly understand and identify the research opportunities behind real-world applications, select the appropriate ML methods to explore, and proactively drive the iterations based on clear analysis of the current results