Machine Learning and AI Safety Research Intern
Confirmed on the employer's own hiring board on Oct 27, 2026. First seen by Alion on Oct 26, 2026. Apple scores A on the Alion truth index.
Are you passionate about shaping the future of safe and responsible artificial intelligence? We are an AI Research and Safety team focused on developing machine learning systems that protect users and deliver remarkably safe experiences across generative AI features. We are seeking a highly motivated and research-driven Machine Learning and AI Safety Research Intern to dive deep into frontier AI technology. In this role, you will focus on applied research in generative AI safety, alignment, and model robustness. You will have the unique opportunity to collaborate with world-class research and engineering teams, transforming your innovative findings into practical, real-world solutions and contributing to impactful research papers. If you thrive on solving ambiguous challenges and want to help build trustworthy AI systems for millions of users, we want to hear from you.
Description
As a Machine Learning and AI Safety Research Intern, you will be at the forefront of AI safety initiatives, working at the intersection of applied research and engineering. Your key responsibilities will include:
Applied AI Research: Conduct cutting-edge applied research focusing on generative AI safety, model alignment, and robustness against adversarial attacks (e.g., prompt injection, jailbreaks, and probing).
Safety Evaluation & Benchmarking: Design, develop, and scale comprehensive LLM evaluation pipelines, safety datasets, and benchmark suites to systematically identify model vulnerabilities.
Infrastructure & Frameworks: Utilize and optimize LLM training frameworks (e.g., DeepSpeed) and high-throughput inference engines (e.g., vLLM, SGLang) to conduct programmatic LLM API calling and batch evaluations.
Cross-Functional Collaboration: Partner closely with multidisciplinary research and engineering teams to translate theoretical safety concepts and guardrails into actionable, deployable engineering solutions.
Academic Contribution: Synthesize your experimental results and safety discoveries to co-author and publish research papers at top-tier AI/ML conferences.
Minimum Qualifications
Currently enrolled in an MS or Ph.D. program in Computer Science, Electrical Engineering, or a related quantitative field.
Hands-on experience with LLM training frameworks (e.g., DeepSpeed) and core machine learning ecosystems (e.g., PyTorch, Hugging Face).
Experience with LLM inference frameworks (e.g., vLLM) and programmatic LLM API calling, including client SDKs and batch evaluation techniques.
Proven experience in designing and building LLM evaluation pipelines, custom datasets, and comprehensive benchmark suites.
Preference for a 5 to 6-month internship duration between Oct 1, 2026, and Sep 30, 2027.
Preferred Qualifications
Strong research experience in AI safety, model alignment, or red-teaming, ideally demonstrated by a track record of publications (or papers currently under review) at premier AI/ML academic venues.
A proactive habit of staying up-to-date with emerging trends and state-of-the-art developments in generative AI and AI agents.
Exceptional critical-thinking, problem-solving, and communication skills, with a proven ability to break down complex, ambiguous open-ended research problems into actionable engineering solutions.
Deep familiarity with safety alignment concepts such as RLHF, guardrail implementation, and adversarial probing.