蚂蚁金服蚂蚁集团-AI Feedback Loop 专家-杭州/上海
任职要求
1.架构设计： 5年以上经验。具备Data-Centric AI 的系统设计能力。能独立设计从用户行为到模型更新的端到端自动化链路。 2.AI 理解力： 需要深刻理解 RLHF, Reward Model, SFT 的原理。深刻理解样本的质量。 3.复杂归因： 在归因分析（Attribution Analysis）和因果推断（Causal Inference）有深入实践，能解决数据噪音和反馈稀疏的问题。 4.…
工作职责
1. 建设高实时性的用户反馈数据流： 负责构建大规模，高并发的实时数据链路，处理海量用户交互日志。 负责对多模态的用户反馈信号进行全链路的清洗，归因与结构化处理，为模型迭代提供标准化的数据养料。 2. 打造自动化的大模型演进闭环 建设面向RLHF和DPO的数据工程体系，自动化构建高质量的偏好数据集，并制成模型的持续对齐与优化。 设计Bad Case自动发现与诊断机制，通过数据驱动的方式识别模型与知识短板，协同算法与知识工程团队形成质量提升的闭环，持续降低模型幻觉率。 3. 质量归因: 解决多轮交互场景下的归因难题：通过因果推断等技术手段，精准定位用户体验波动的核心因子，知道算法迭代方向。
1. 建设高实时性的用户反馈数据流： 负责构建大规模，高并发的实时数据链路，处理海量用户交互日志。 负责对多模态的用户反馈信号进行全链路的清洗，归因与结构化处理，为模型迭代提供标准化的数据养料。 2. 打造自动化的大模型演进闭环 建设面向RLHF和DPO的数据工程体系，自动化构建高质量的偏好数据集，并制成模型的持续对齐与优化。 设计Bad Case自动发现与诊断机制，通过数据驱动的方式识别模型与知识短板，协同算法与知识工程团队形成质量提升的闭环，持续降低模型幻觉率。 3. 质量归因: 解决多轮交互场景下的归因难题：通过因果推断等技术手段，精准定位用户体验波动的核心因子，知道算法迭代方向。
Key Responsibilities Onboarding and training: Guide new clients through setup, ensuring smooth adoption and confidence in using Evelab Insight tools. Relationship management: Build strong, long-term partnerships with clients, acting as their advocate within Evelab Insight. Pain Point Identification: Identify client pain points, design and develop tailored solutions that maximize efficiency and ROI. Solution optimization: Track client usage and outcomes, proactively addressing risks and surfacing insights. Upsell and expansion: Spot opportunities to introduce new features or services aligned with client goals. Cross-functional collaboration: Partner with Sales, Product, and Support teams to deliver seamless client experiences. Feedback loop: Capture client insights to inform product development and service improvements. Required
● Lead the FDE team of Alibaba Cloud International Business to partner with key customers to design, co-develop, debug and deploy vertical agentic solutions that generate measurable impact. ● Provide technical guidance on model tuning, agent workflows, CI/CD pipelines, and production readiness, ensuring rapid iteration from prototype to robust, scalable systems. ● Drive the evolution of the Agent Development Platform, working with internal engineer teams to create reusable tools, documentation, and accelerators, enabling developers to easily build, test and deploy agents. ● Systematize field insights from customer engagements and share learnings with product teams to improve future experiences and capabilities. ● Establish a structured feedback loop with product and engineering teams, synthesizing field insights from customer engagements to improve future experiences and capabilities. ● Elevate engineering excellence across the FDE team by establishing code standards, architectural best practices and benchmarks.