Member of Technical Staff, Research — Early Career(PHD

About us

We are building AI systems that can reason, use tools, and complete meaningful work in the real world. Our team works across model post-training, reinforcement-learning infrastructure, large-scale training, and product engineering. We believe the fastest path to more capable and reliable agents is an integrated loop: challenging environments, rigorous evaluations, efficient training, reliable inference, and products that make those capabilities useful.

You will own research at the intersection of agentic reinforcement learning, post-training, evaluation, environments, and scaling. You will identify high-leverage questions, design and run decisive experiments, and translate results into model improvements and reusable systems.

This role is designed for researchers completing or recently completing a PhD, as well as candidates with an equivalent record of original research. It is not a purely academic position: strong candidates write excellent code, work closely with systems engineers, and care whether an idea survives realistic evaluation and production constraints.

• Formulate high-leverage research questions about agentic capability and reliability, RL algorithms, reward and verifier design, exploration, curricula, environment design, task distributions, and scaling behavior.

• Design rigorous experiments, ablations, controls, and evaluations that separate real model improvement from noise, data leakage, reward hacking, or benchmark overfitting.

• Implement new methods in modern deep-learning frameworks and integrate them with production training, rollout, environment, and evaluation systems.

• Build or improve datasets, agent environments, verifiers, and evaluations for domains such as coding, tool use, reasoning, long-horizon tasks, or computer interaction.

• Analyze trajectories and model behavior, develop useful failure taxonomies, and turn observations into testable hypotheses and prioritized experiments.

• Partner closely with Post-Training, RL Systems, Training, and Backend & Product engineers to scale promising ideas and expose them to realistic product constraints.

• Communicate findings in clear internal documents and technical reviews; contribute to papers, technical reports, blog posts, or open-source releases when aligned with company goals.

• Help shape the research roadmap by identifying compounding capabilities, reusable evaluation assets, and experiments that retire the most important uncertainties.

• Completing or recently completed a PhD in computer science, machine learning, statistics, mathematics, or a related field—or an equivalent record of original, technically rigorous research.

• A strong research record in machine learning, reinforcement learning, large language models, agents, or ML systems, demonstrated through publications, preprints, open-source work, or substantial independent projects.

• Excellent Python skills and hands-on experience with a modern deep-learning framework such as PyTorch or JAX.

• Experimental rigor: you can define a falsifiable question, build the right measurement, control confounders, interpret noisy results, and communicate uncertainty honestly.

• The engineering ability to navigate an unfamiliar codebase, build reliable research infrastructure, and turn a promising idea into a working system.

• Clear written and verbal communication and the ability to collaborate across research, systems, and product disciplines in a fast-moving environment.

• Research experience in agentic reinforcement learning, post-training, preference learning, reward or verifier modeling, evaluation, or environment design.

• Experience with large-model training, distributed inference, high-throughput rollout systems, or performance-sensitive ML infrastructure.

• Notable publications, open-source contributions, datasets, benchmarks, or research artifacts that other people use.

• Domain expertise in coding agents, mathematical reasoning, scientific discovery, tool use, long-horizon planning, or computer interaction.

• Experience transferring a research result into a production model, product, or dependable shared system.

• Mission…