About the roleYou 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.
What you'll do- 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.
Why GoalyWork on frontier AI problems across model training, inference, agentic RL infra, and domain-specialized continual learning.
Build systems that push research into production - from new mode recipes and post-training methods to infrastructure that runs at serious scale.
Publish and contribute to open source and top AI conferences, with opportunities to pursue work worthy of top AI conferences and rele
ase impactful OSS used by the broader AI community.
Serious resources to build with: well-capitalized, substantial GPU capacity, and competitive compensation.
Learn directly from a founding team that has trained trillion-parameter models and built frontier-scale AI infrastructure, while developing your own research and technical leadership.
Own meaningful problems from day one. On a small, highly technical team, you'll have unusually large scope, research freedom, and direct impact on both the product & technical roadmap.
Move fast and own the full problem, not one tiny component. Work directly with the founding team, make technical decisions quickly, and take ideas from research to production without layers of process.
H-1B sponsorship available; OPT/CPT candidates welcome.You may be a good fit if you have- 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.
Strong pluses- 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.
Location, visa sponsorship & benefits- Hybrid in Palo Alto: 4+ days/week in office.
- Visa sponsorship: H-1B and OPT/CPT support available, with immigration counsel.
- Meals & perks: Complimentary lunch, dinner, snacks, and drinks.
A note on qualifications. We value exceptional ability over perfect keyword matches. If the work excites you and you can show strong technical ability, learning speed, or ownership, we encourage you to apply.
