代理工程师
Agent Engineer
职位描述
这个职位位于代理系统、平台工程、集成和产品开发的交汇点，是一家早期阶段AI生产力初创公司正在打造的自主执行助理。你将帮助定义助理的功能、这些功能的工作方式，以及它们如何成为用户可以信赖的真实工作体验。这是一个高度自主的职位，对技术方向和产品成果有直接影响力。
你将负责：
- 研究推理、规划、记忆、工具使用和代理协作的最新进展，以识别有价值的产品机会。
- 将有前景的模型行为转化为可靠、可重用的技能和工作流程，解决真实用户的痛点。
- 构建和演进一个自定义的Python代理框架，包括执行循环、编排、上下文管理、结构化输出、重试和错误恢复。
- 设计跨邮件、日历、消息平台、浏览器、文档、CRM和其他企业软件的代理工具和集成。
- 全面负责Python代理、Django服务、React界面、数据模型、后台任务、可观测性和生产运维的能力。
- 构建平台抽象，使能力在产品变得更复杂时更容易组合、扩展和维护。
- 提高高吞吐量代理执行的延迟、成本、可靠性和安全性。
- 与代理评估团队合作，定义预期行为，实施能力，并将质量发现转化为工程改进。
- 研究生产追踪和用户反馈，了解信任为何失效，然后修复底层系统。
- 在模糊问题上设定技术方向，并通过设计评审和周到的执行提升工程标准。
我们寻找：
- 5年以上从头到尾构建生产软件系统的经验，包括设计、实现、部署和运营迭代。
- 强大的Python技能，能够编写可维护的生产代码，调试异步系统，并设计合理的抽象。
- 实际构建代理系统的经验：规划、工具调用、结构化输出、上下文管理、状态管理、重试和编排。
- 全栈灵活性，具备深厚的Python和Django专业知识，同时熟悉API、React和TypeScript。
- 设计和构建与外部系统集成的经验
查看英文原文
ABOUT THE ROLE
This role sits at the intersection of agent systems, platform engineering, integrations, and product development at an early-stage AI productivity startup. You will define and build the core capabilities of an AI assistant that executives and founders rely on for real, high-stakes work. Your work directly shapes what the product can do and how dependably it does it.
WHAT YOU'LL DO
- Investigate advances in reasoning, planning, memory, tool use, and agent collaboration to identify valuable product opportunities.
- Translate promising model behaviors into reliable, reusable skills and workflows that solve real user problems.
- Build and evolve a custom Python agent harness, including execution loops, orchestration, context management, structured outputs, retries, and error recovery.
- Design agent tools and integrations across email, calendars, messaging platforms, browsers, documents, CRMs, and business software.
- Own capabilities end-to-end across the Python agent, Django services, React interfaces, data models, background jobs, observability, and production operations.
- Build platform abstractions that make capabilities easier to compose, extend, and maintain as the product grows more sophisticated.
- Improve latency, cost, reliability, and safety across high-volume agent execution.
- Partner with the Agent Evaluations team to define expected behavior, instrument capabilities, and turn quality findings into engineering improvements.
- Study production traces and user feedback to understand where users lose trust, then fix the underlying system.
- Set technical direction on ambiguous problems and raise the engineering standard through design reviews and thoughtful execution.
WHAT WE'RE LOOKING FOR
- 5 or more years building production software systems end-to-end, including design, implementation, deployment, and operational iteration.
- Strong Python proficiency: maintainable production code, asynchronous systems, and sound abstractions.
- Hands-on experience building agent systems, including planning, tool calling, structured outputs, context management, state management, retries, and orchestration.
- Full-stack capability with deep Python and Django expertise, plus comfort working with APIs, React, and TypeScript.
- Experience designing and building integrations with external APIs and business software platforms such as email, calendars, messaging, and CRMs.
- Experience designing reusable platform abstractions that enable composition, extension, and maintenance at scale.
- Analytical debugging ability across prompts, traces, model outputs, application code, databases, and user interactions.
- Product judgment to turn vague user needs and emerging technical possibilities into simple, useful capabilities without requiring complete specifications.
- Strong CS fundamentals, ideally from an engineering-focused academic background.
- Prior startup experience or demonstrated career progression with increasing ownership within a single organization.
- Familiarity with LLM-based systems, prompt engineering, or model evaluation in production is a plus.
- Experience with agent frameworks such as LangChain or AutoGen is a plus.
- Background in workflow automation, autonomous systems, or distributed systems optimization is a plus.
LOCATION
On-site in Palo Alto, CA with hybrid flexibility. Visa sponsorship is not available for this role.