Applied Research - Forward-Deployed

Responsibilities

• Embed with strategic customers to understand agent architectures, failure modes, and product goals.
• Design custom reinforcement-learning environments, evaluation harnesses, verifiers, reward functions, and rubrics.
• Architect agent scaffolding for tool use, multi-step reasoning, memory, and sandbox execution.
• Configure and launch training runs on the Lab platform, iterating on rollout strategies and evaluation criteria.
• Lead technical engagements from discovery through deployed and improved models.
• Identify repeatable customer patterns and codify them into reference implementations, templates, documentation, and recipes.
• Provide customer feedback to shape the platform roadmap.
• Build examples and contribute to technical content, including blog posts, tutorials, and case studies.
• Develop evaluation methods for agentic behavior, including reasoning, tool-use correctness, failure recovery, and long-horizon task completion.
• Prototype agent harnesses for code generation, workflow automation, document processing, and other real-world tasks.
• Stay current on agentic AI, evaluations, and post-training methods.

Requirements

• Hands-on experience building, evaluating, or deploying LLM-based agents.
• Strong evaluation-design skills, including measurement, rubric construction, and reward-signal assessment.
• Working knowledge of reinforcement learning and post-training concepts, including GRPO, RLHF, reward modeling, and SFT.
• Strong Python skills and familiarity with modern AI tooling.
• Experience with Hugging Face, inference engines, and agent frameworks.
• Experience in a customer-facing or consulting-oriented technical role, or as a technical founder.
• Excellent written and verbal communication skills.
• High agency and comfort working in ambiguous environments.

Nice-to-Haves

• Experience with DSPy, LangGraph, MCP, Stagehand, or browser automation.
• Experience building or operating LLM evaluation pipelines at scale, including benchmarks, synthetic data generation, and model grading.
• Research experience, publications, open-source contributions, or ML/RL/agent benchmarks.
• Familiarity with sandbox and code-execution environments for agent evaluation.
• Web programming experience with React, TypeScript, or Next.js.

Compensation and Benefits

• Cash compensation range: $150,000–$300,000, plus equity incentives.
• Flexible work options, including San Francisco or hybrid-remote arrangements.
• Visa sponsorship and relocation support.
• Professional development budget.
• Team off-sites and conference attendance.