Key Responsibilities
1. Participate in the migration from Dify to LangGraph agent framework
- Understexisting Dify workflows, agents tools
- Re-design implement equivalent (improved) workflows using LangGraph
2. Develop maintain Python-based AI agents, including:
- Agentchestration (planner / execut/ tool-calling patterns)
- Multi-step task execution state management
3. Design optimize context construction prompt engineering fLLM-based agents
4. Integrate agents with existing services (e.g. document parsing, retrieval, rule checking, APIs)
5. Collaborate with internal tech team on architecture alignment, code review best practices
4. Produce clear technical documentation fagent workflows framework migration.
Required Skills & Experience
1. Solid Python programming skills, with clean maintainable coding practices
2. Hands-on experience with Dify (workflow design, agent configuration, tools, prompts)
3. Hands-on experience with LangGraph, including:
- Graph-based agentchestration
- State, memory conditional routing
4. Strong understanding of Agentic AI concepts, such as:
- Agentchestration
- Tool calling
- Multi-agent multi-step reasoning workflows
5. Strong experience in prompt engineering & context engineering
6. Familiar with LLM-based application development (Qwen/ AWS Bedrock / Azure OpenAI / similar)
Preferred (Nice to Have)
1. Experience with RAG pipelines (document retrieval, chunking, embedding)
2. Familiarity with enterprise AI considerations (logging, traceability, errhandling)
3. Experience in healthcare / pharma / compliance-related AI projects is a plus

主要職責(zé)
1. 參與從 Dify 遷移到 LangGraph 代理框架的過程
- 了解現(xiàn)有的Dify工作流程、代理和工具
- 重新設(shè)計和使用LangGraph實現(xiàn)等效（或改進）的工作流程
2. 開發(fā)和維護基于Python的AI代理，包括：
- 代理編排（規(guī)劃器/執(zhí)行者/工具調(diào)用模式）
- 多步驟任務(wù)執(zhí)行與狀態(tài)管理
3. 設(shè)計和優(yōu)化基于LLM的代理的上下文構(gòu)建與提示工程
4. 將代理與現(xiàn)有服務(wù)（如文檔解析、檢索、規(guī)則檢查、API）集成
5. 與內(nèi)部技術(shù)團隊合作，進行架構(gòu)對齊、代碼審查和實踐
4. 為代理工作流程和框架遷移制作清晰的技術(shù)文檔。
必備技能與經(jīng)驗
1. 扎實的Python編程技能，編碼流程干凈且易于維護
2. Dify的實際操作體驗（工作流程設(shè)計、代理配置、工具、提示）
3. LangGraph的實際操作經(jīng)驗，包括：
- 基于圖的代理編排
- 狀態(tài)、內(nèi)存和條件路由
4. 對智能人工智能概念的深入理解，例如：
- 代理編排
- 工具調(diào)用
- 多代理或多步驟推理工作流程
5. 在提示工程和上下文工程方面有扎實的經(jīng)驗
6. 熟悉基于大型語言模型的應(yīng)用開發(fā)（Qwen/AWS Bedrock/Azure OpenAI 等）
優(yōu)選（擁有就好）
1. RAG管道經(jīng)驗（文檔檢索、分塊、嵌入）
2. 熟悉企業(yè)AI相關(guān)考量（日志記錄、可追溯性、錯誤處理）
3. 有醫(yī)療/制藥/合規(guī)相關(guān)AI項目經(jīng)驗者加分