阿里巴巴数据技术及产品部-大模型RL Data工程师-长程 Agent 环境(RL 数据)
社招全职3年以上地点:杭州状态:招聘
工作描述
任职要求 1. 对长程多步 agentic 任务及其失败模式有深入判断,能将真实复杂任务转化为可复现、可自动判分、难度可控且抗作弊的训练与评测环境;有 Agent 产品实测、长程任务或轨迹设计、过程级评测、RL 环境或数据经验者优先。 2. 具备很强的抽象与系统化能力,能将长任务拆解为子目标与里程碑 checkpoint,设计难度分级与变体生成策略,并把模糊的「做得好不好」落成可自动验证的过程评分,沉淀为可跨垂直复用的方法论。 3. 工程能力扎实,熟练掌握 Python(能独立编写判分器 verifier、解析长轨迹、处理大规模 rollout 数据),熟悉至少一类 agent 执行环境(终端 / shell、浏览器自动化、桌面自动化、代码沙箱或 MCP 工具编排),掌握以 mock / 录制替代真实依赖、容器化(Docker)与环境状态的重置 / 回放 / 分支。 4. 了解 Agent 核心机制(ReAct、Tool Use、Planning、多步推理与反思)与后训练范式(SFT / RLHF / 可验证奖励 RL、RLVR),理解过程监督(step-level reward)、跨步 credit assignment、reward hacking 与长上下文下的误差累积,对长程 agentic 任务的能力边界有认知或实践。 5. 领域呈 T 型即可,熟悉任一垂直(编程、数据分析、办公、科研、金融等)到能判断长任务是否贴近真实,但核心能力在长程环境与过程奖励工程、服务所有方向;有以下经验者优先:agent 评测框架 / RL 环境搭建、benchmark / 任务集构建、过程级 verifier 与自动评分器开发、大规模 roll…
登录查看完整工作描述
微信扫码,1秒登录
包括英文材料
AI agent+
https://www.ibm.com/think/ai-agents
Your one-stop resource for gaining in-depth knowledge and hands-on applications of AI agents.
Python+
https://liaoxuefeng.com/books/python/introduction/index.html
中文,免费,零起点,完整示例,基于最新的Python 3版本。
https://www.learnpython.org/
a free interactive Python tutorial for people who want to learn Python, fast.
https://www.youtube.com/watch?v=K5KVEU3aaeQ
Master Python from scratch 🚀 No fluff—just clear, practical coding skills to kickstart your journey!
https://www.youtube.com/watch?v=rfscVS0vtbw
This course will give you a full introduction into all of the core concepts in python.
Bash+
[英文] The Bash Guide
https://guide.bash.academy/
A quality-driven guide through the shell's many features.
https://www.youtube.com/watch?v=tK9Oc6AEnR4
Understanding how to use bash scripting will enhance your productivity by automating tasks, streamlining processes, and making your workflow more efficient.
MCP+
https://www.youtube.com/watch?v=eur8dUO9mvE
Unlock the secrets of MCP! 🚀 Dive into the world of Model Context Protocol and learn how to seamlessly connect AI agents to databases, APIs, and more. Roy Derks breaks down its components, from hosts to servers, and showcases real-world applications. Gain the knowledge to revolutionize your AI projects!
https://www.youtube.com/watch?v=L94WBLL0KjY
Let's talk about MCP or the Model Context Protocol.
Docker+
https://www.youtube.com/watch?v=GFgJkfScVNU
Master Docker in one course; learn about images and containers on Docker Hub, running multiple containers with Docker Compose, automating workflows with Docker Compose Watch, and much more. 🐳
https://www.youtube.com/watch?v=kTp5xUtcalw
Learn how to use Docker and Kubernetes in this complete hand-on course for beginners.
React+
[英文] Quick Start - React
https://react.dev/learn
This page will give you an introduction to 80% of the React concepts that you will use on a daily basis.
https://www.youtube.com/watch?v=SqcY0GlETPk
Master React 18 with TypeScript! ⚛️ Build amazing front-end apps with this beginner-friendly tutorial.
https://www.youtube.com/watch?v=x4rFhThSX04
Learn modern React basics in the most interactive, hands-on way possible in the full course for beginners.
还有更多 •••
相关职位
社招3年以上
1. 金融、会计、财务、审计、税务、经济、金融工程或相关专业本科及以上学历,对金融财会领域大模型的能力边界有判断,能将真实财务任务转化为可标注、可评测的数据;持有 CPA / CFA / ACCA /
更新于 2026-09-22杭州
社招2年以上技术类-算法
1、硕士及以上学历,计算机科学与技术、数学等相关专业优先; 2、有2年以上大模型相关算法研发经验,熟悉常见的深度学习框架(如Tensorflow、Pytorch),有Agentic大模型开发经验优先;
更新于 2026-09-07北京|杭州
社招3年以上
1.语音 / 音频领域 3 年以上研发经验(高质量博士 / 硕士研究阶段计入),在所选方向至少一个细分领域有深度技术积累;有论文发表记录优先; 2.具备评测系统设计和实施经验,能独立将模糊的质量维度转
更新于 2026-07-20北京|杭州