Research Scientist (AI)

TL;DR

Research Scientist (AI): Developing and applying post-training methods and interpretability techniques to enhance the safety and robustness of frontier AI systems with an accent on RLHF, DPO, and GRPO methodologies. Focus on designing post-training pipelines, conducting interpretability-informed evaluations, and translating research findings into actionable safety standards for industry and policymakers.

Location: Must be based in the United States

Company

hirify.global is a leading data and evaluation partner for frontier AI companies, providing high-quality data and full-stack technologies to help enterprises and governments build, deploy, and oversee impactful AI applications.

What you will do

• Design and execute post-training pipelines to analyze the impact of training choices on model safety, robustness, and alignment.
• Develop interpretability-informed evaluations to identify and mitigate unsafe, deceptive, or undesirable model behaviors.
• Collaborate with cross-functional teams, including policymakers and engineers, to translate research into actionable safety standards and benchmarks.
• Conduct rigorous research on agent robustness, AI control protocols, and risk evaluation.
• Publish research findings to contribute to the broader scientific understanding of AI capabilities and risks.

Requirements

• At least three years of experience addressing sophisticated machine learning problems in research or product development.
• Proven experience with post-training and RL techniques such as RLHF, DPO, or GRPO.
• Strong track record of published research in machine learning, specifically within generative AI.
• Excellent written and verbal communication skills for cross-functional collaboration.
• Commitment to the mission of promoting safe, secure, and trustworthy AI deployments.

Nice to have

• Experience with mechanistic interpretability, probing, or understanding model internals.
• Familiarity with red-teaming or adversarial evaluation of post-trained models.
• Experience studying failure modes like reward hacking, sycophancy, or alignment faking.

Culture & Benefits

• Comprehensive health, dental, and vision coverage.
• Retirement benefits and equity-based compensation.
• Generous paid time off (PTO) and learning/development stipends.
• Inclusive and equal opportunity workplace culture.
• Opportunities to collaborate with industry leaders and government agencies.

Hiring process

• Interviews focus on practical ML prototyping, debugging, and research concept grasp.
• No LeetCode-style questions are used in the evaluation process.