3 hours ago
Base Salary
$180k - $400k/yr
Responsibilities
- Build and scale interactive computer-use environments, task generators, stateful workflows, success checks, and synthetic training trajectories.
- Design benchmarks and evaluation harnesses for task success, reliability, efficiency, and safety, and convert model failures into training or environment improvements.
- Develop and scale supervised fine-tuning, reinforcement learning, reward modeling, graders, and verifier techniques for long-horizon agents.
- Research, prototype, and rigorously test novel algorithms using strong baselines and ablations.
- Build data filtering, curation, difficulty calibration, and curriculum methods for high-signal training runs.
- Build and scale training and inference infrastructure for agent rollouts and RL while improving throughput, latency, reproducibility, observability, and cost.
- Collaborate with harness, product, and infrastructure teams to improve shipped models and debug difficult production failures.
Requirements
- Strong software engineering and prototyping skills with experience building reliable systems used by researchers and engineers.
- Demonstrated research ability, including forming hypotheses, designing rigorous experiments, and interpreting noisy results.
- Hands-on experience training, fine-tuning, or evaluating large language or multimodal models.
- Deep expertise in at least one of data and environment engineering, agentic evaluations, RL post-training, or training and inference infrastructure for agents.
- Ability to work independently through ambiguity, take ownership, and pursue promising technical ideas.
- Clear technical communication and ability to collaborate across research, infrastructure, and product teams.
- Preferred experience includes computer-use, coding, web, browser, or tool-using agents; long-horizon or sparse-reward RL; reward models, LLM judges, or verifiers; large-scale distributed training or rollout systems; relevant publications or open-source work; and applying product signals to improve shipped AI products.
Categories
AI ResearchML Engineering
