Research Scientist Intern PhD, PAR Safety Evals and Red Teaming

Meta; Menlo Park, New York (CA, NY); On-site; USD 55,000 - 69,000; Full time

Job summary

Meta is seeking Research Interns to join Safety Evals and Red Teaming within Products and Applied Research. You will work on agent safety, benchmarks, synthetic data, automated red teaming, and LLM-based judging.

Internships are 16 weeks with multiple start dates. You’ll design experiments, analyze data, and communicate results to both technical and non-technical audiences, using Python, C++, and ML tools.

Qualifications

• PhD in Computer Science, AI, or related field, or currently pursuing with expected completion.
• Experience in safety, red-teaming, evals, ML or NLP research.
• Proficiency in Python, C++, Java or similar languages; strong experimental design skills.

Responsibilities

• Design and run ML safety experiments and red-teaming evaluations.
• Analyze results, synthesize findings, and communicate insights to stakeholders.

Skills

Python

C++

ML/DL

Research

Technical writing

Education

PhD in CS/AI/related field

Tools

ML frameworks

Git

Job description

Meta was built to help people connect and share, and over the last decade our tools have played a critical part in changing how people around the world communicate with one another. In the pursuit of personal superintelligence, Meta's AI systems are becoming more capable, and evaluating their safety is critical.

Meta is seeking Research Interns to join our Safety Evals and Red Teaming team within the Products and Applied Research org.

We are seeking individuals passionate in areas such as Agent Safety, Benchmarks and Evaluations, Synthetic Data Generation, Automated Red teaming, LLM-as-a-Judge, benchmark design, eval harnesses, prompt injection, agent environments.

Our internships are 16 weeks long and we have various start dates throughout the year.

What You Could Work On:

• How reliably do LLM-based judges identify safety-policy violations across different tasks, prompts, and adversarial conditions? How should we measure their errors and consistency?
• How should we evaluate prompt-injection risks when an agent interacts with tools and untrusted information? Which environments and attack scenarios expose failures that simpler tests miss?
• Which automated red teaming methods uncover meaningful safety failures, and how well do those failures generalize beyond the prompts or environments used to find them?
• How can we use synthetic data and benchmark design to improve coverage while checking that scores reflect the behavior we intend to measure?

• Currently has or is in the process of obtaining a Ph.D. degree in Computer Science, Computer Vision, Audio Processing, Artificial Intelligence, Generative AI, or relevant technical field
• Research experience in safety, red-teaming, evals, machine learning, deep learning, computer vision, and/or natural language processing
• Experience with Python, C++, C, Java or other related languages
• Must obtain work authorization in the country of employment at the time of hire and maintain ongoing work authorization during employment
• Experience in one or more areas relevant to the team: model evaluation, LLM safety or alignment, agent security, prompt injection, automated red teaming, synthetic data generation, LLM-based judging, or evaluation science
• Experience designing ML experiments, analyzing complex datasets, and using quantitative approaches to investigate research questions
• Substantial research contributions through first-author publications at leading workshops or conferences such as NeurIPS, ICLR, AAAI, RecSys, KDD, IJCAI, CVPR, ECCV, ACL, NAACL, EACL, ICASSP, or similar
• Demonstrated software engineering experience through internships, employment, coding competitions, or contributions to widely used open-source projects
• Experience collaborating with teams and communicating technical results to non-technical audiences
• Intent to return to the degree program after the completion of the internship/co-op