AI Architect
Description
Company Overview
Yuwell-POCTech is the glucose and metabolic monitoring division of Yuwell Group, dedicated to building world-class CGM hardware and digital chronic disease management platforms. Our Anytime series CGM is MDR-certified and serves users across more than 100 countries and regions.
Beyond hardware innovation, we are building "AnySugar" and next-generation user-facing AI Agents to deliver personalized, conversational metabolic health coaching at scale.
Position Summary
We are seeking an experienced AI Architect to lead the design and evolution of our user-facing Agent systems. This role focuses on creating intelligent, conversational health companions that transform CGM data into actionable, personalized care through natural language interaction and proactive intervention.
Key Responsibilities
- Agent System Architecture
- Architect the core AI Agent system for our consumer health platform, including intent recognition, task planning, tool use, memory management, and personalization.
- Design multi-Agent collaboration frameworks for scenarios such as diabetes education, diet coaching, exercise guidance, and medication adherence.
- Conversational AI & Content Generation
- Build multi-turn dialogue management, contextual understanding, sentiment analysis, and natural language generation capabilities.
- Design multimodal output formats including graphical reports, voice broadcasting, and health insight notifications.
- Personalized Health Reasoning
- Develop dynamic health profiles by integrating calibrated CGM data with user inputs (diet, activity, sleep, medication).
- Drive Agent-generated personalized recommendations for diet, exercise, risk alerts, and behavioral interventions.
- RAG & Medical Knowledge Engineering
- Establish Retrieval-Augmented Generation (RAG) pipelines with medical knowledge bases (clinical guidelines, nutrition science).
- Ensure output accuracy, compliance, and implement hallucination mitigation strategies.
- Agent Platform & Tool Ecosystem
- Design Agent orchestration frameworks integrating with "AnySugar" business systems, external APIs, and wearable ecosystems.
- Establish prompt engineering, version control, and A/B testing mechanisms for iterative improvement.
- Evaluation, Safety & Compliance
- Build automated evaluation systems covering factuality, safety, user satisfaction, and task completion rates.
- Ensure data privacy, content safety, and regulatory compliance in healthcare contexts.
- Technology Strategy
- Track advancements in LLMs, Agent architectures, and multimodal models.
- Explore innovative applications such as digital twins, virtual health coaching, and family health collaboration.
Requirements
- Master's degree or above in Computer Science, AI, NLP, HCI, or Biomedical Informatics; PhD preferred.
- 3+ years of experience in AI/ML application development; 1+ year in Agent system architecture design.
- Deep understanding of LLM principles and hands-on experience with Agent frameworks (LangGraph, AutoGen, Dify, CrewAI, etc.).
- Proven experience in RAG, knowledge graphs, prompt engineering, and function calling/tool use.
- Track record in building conversational AI, chatbots, or intelligent assistant platforms.
- Proficiency in Python, cloud-native architecture, vector databases (Milvus, Pinecone, etc.), and message queues.
- Excellent cross-functional collaboration skills, able to work effectively with product, engineering, medical, and operations teams.
Preferred Qualifications
- Experience in digital health, DTx, or medical AI conversational products.
- Publications or patents in NLP, Agent, or medical AI fields.
- Familiarity with healthcare data compliance (HIPAA, GDPR, NMPA).
- Proven success in building and scaling C-end AI assistants from scratch.
What We Offer
- Real-world impact: Architect AI Agents for millions of CGM users globally.
- Rich data assets: Multimodal health signals at scale.
- Global expansion: MDR-certified products in APAC, Europe, LATAM, MENA.
- Competitive compensation, equity options, and collaboration opportunities with leading medical experts.
📍 Location: Nanjing / Shanghai (hybrid)
鱼跃凯立特是鱼跃集团旗下专注血糖及代谢监测的子公司,业务覆盖 BGM、CGM、POCT 与数字化慢病管理全链路。我们自主研发的 Anytime 系列 CGM 产品已获欧盟 MDR 认证,业务覆盖全球 100 余个国家和地区。
在硬件研发之外,我们正全力构建“安耐糖”等用户端数字化平台,致力于打造面向 C 端用户的代谢健康管理 AI Agent。我们期望通过自然语言交互、个性化建议和主动式干预,将千万级血糖数据转化为切实的用户健康价值,实现从“监测”到“陪伴与干预”的跨越。
岗位名称
AI 架构师
岗位使命
负责鱼跃凯立特用户端 AI Agent 系统的总体架构设计与技术演进,打造“主动感知、自然交互、专业可靠、持续进化”的健康管理智能体,连接 CGM 数据与用户日常行为,构建从监测到干预的完整数字化闭环。
核心职责
- 用户端 Agent 系统架构设计
- 主导面向糖尿病及代谢健康用户的 AI Agent 核心架构,涵盖意图理解、任务规划(Planning)、工具调用(Tool Use)、记忆机制(Memory)与个性化画像体系。
- 设计多 Agent 协作框架,支持健康咨询、疾病教育、饮食建议、运动指导、用药提醒等多场景业务协同。
- 智能交互与内容生成
- 构建多轮对话管理、上下文理解、情感识别与自然语言生成能力,确保交互体验的专业性与连贯性。
- 设计图文报告、语音播报、健康洞察推送等多模态输出形态,适配不同用户偏好。
- 个性化健康推理引擎
- 基于已校准的 CGM 血糖数据,融合饮食、运动、用药、睡眠等多模态用户输入,构建动态健康画像。
- 驱动 Agent 生成个性化饮食推荐、运动计划、风险预警与行为干预建议,提升用户依从性。
- RAG 与医学知识工程
- 建设糖尿病管理、临床指南、营养学知识的检索增强生成(RAG)体系,对接权威医学知识库。
- 确保 Agent 输出的专业性与合规性,建立知识更新与幻觉抑制机制。
- Agent 平台与工具生态
- 设计 Agent 编排框架,对接“安耐糖”业务系统、外部食谱/运动 API 及可穿戴设备生态。
- 建立 Prompt 工程体系、Few-shot 管理、版本控制与 A/B 实验机制,支撑业务快速迭代。
- 评估、安全与合规
- 建立 Agent 自动化评估体系,涵盖事实一致性、安全性、用户满意度及任务完成率等核心指标。
- 保障医疗健康场景下的数据隐私、内容安全与合规审计,符合相关法规要求。
- 技术战略与前沿探索
- 跟踪大语言模型(LLM)、Agent 架构、多模态大模型等前沿技术方向。
- 探索数字孪生、虚拟健康教练、家庭多成员健康协作等创新产品形态。
任职要求
- 硕士及以上学历,计算机科学、人工智能、自然语言处理、人机交互、生物医学信息学等相关专业;博士优先。
- 3 年以上 AI/ML 应用开发经验,1 年以上 Agent 系统架构设计经验。
- 深入理解大语言模型(LLM)原理,熟悉主流 Agent 框架(如 LangGraph、AutoGen、Dify、CrewAI 等)。
- 具备 RAG、知识图谱、Prompt Engineering、Function Calling / Tool Use 的实战经验。
- 具备多轮对话系统、智能助手或 Chatbot 平台的设计与落地经验。
- 熟悉 Python、云原生架构、向量数据库(如 Milvus、Pinecone 等)及消息队列技术。
- 具备优秀的跨团队协作能力,能高效联动产品、前端、后端、医学及运营团队,交付高质量的 C 端产品体验。
加分项
- 具备健康管理、数字疗法(DTx)、医疗 AI 对话产品相关经验。
- 发表过 NLP、Agent 或医疗 AI 领域的学术论文或持有相关专利。
- 熟悉医疗数据合规规范(如 HIPAA、GDPR 或 NMPA 相关要求)。
- 有从 0 到 1 打造 C 端 AI 助手并实现规模化落地的成功经验。
我们提供
- 真实的 C 端应用场景:千万级 CGM 用户入口及亿级交互频次的健康管理平台。
- 丰富的数据资产:血糖、饮食、运动、用药等多模态真实世界数据。
- 全球化视野:已获认证的出海产品矩阵,提供面向全球用户的 Agent 本地化机会。
- 有竞争力的薪酬福利、期权及项目激励机制,以及与顶级医学专家共建 AI 健康助手的发展平台。
📍 工作地点:南京 / 上海(可协商)
Skills
Sign in to read and join the discussion.