人工智能应用工程师 - LangGraph & 代理AI
AI Application Engineer - LangGraph & Agentic AI
我们正在寻找一位经验丰富的AI应用工程师，负责设计和构建由大语言模型（LLM）、LangGraph和现代代理AI技术驱动的智能应用。
你将专注于将业务需求转化为能够通过任务推理、信息检索、与工具和企业系统交互、在需要时请求人工审批并完成业务流程的AI应用。
该职位位于AI工程、软件开发、流程自动化和业务流程转型的交汇点。
要求
代理AI应用开发
- 使用LangGraph和LLM技术设计和开发AI应用。
- 构建能够执行复杂多步骤业务流程的代理。
- 设计包含推理、工具使用、验证、审批和异常处理的状态化流程。
- 在适当的情况下开发单代理和多代理解决方案。
- 将业务需求转化为实际的代理AI架构。
LLM应用工程
- 将LLM集成到生产级应用中。
- 开发提示策略、结构化输出、工具调用和上下文管理方法。
- 根据准确性、能力、延迟、安全性和成本选择合适的模型。
- 开发提高可靠性和减少幻觉的机制。
- 在AI生成的决策和操作周围实施适当的约束。
RAG与企业知识
- 设计和实现检索增强生成（RAG）解决方案。
- 将AI应用连接到企业文档、数据库、API和知识库。
- 开发检索和排序策略，为代理提供相关上下文。
- 使用嵌入和向量数据库。
- 实现支持AI代理的数据和上下文管道。
业务流程自动化
- 分析业务流程并识别代理自动化的机遇。
- 设计结合LLM推理和确定性业务逻辑的AI流程。
- 构建能够检索信息、做出决策、调用工具并完成操作的代理。
- 实现人机协作的审批和升级流程。
- 确保自动化操作可控、可审计并在适当情况下可逆。
评估与质量
- 为AI应用和代理流程开发评估框架。
- 定义涵盖准确性、任务完成率、可靠性、延迟和成本的指标。
- 构建针对提示工程的自动化测试。
查看英文原文
We are looking for an experienced AI Application Engineer to design and build intelligent applications powered by LLMs, LangGraph, and modern agentic AI technologies.
You will focus on transforming business requirements into AI applications capable of reasoning through tasks, retrieving information, interacting with tools and enterprise systems, requesting human approval when required, and completing business processes.
This role sits at the intersection of AI engineering, software development, workflow automation, and business process transformation.
Requirements
Agentic AI Application Development
- Design and develop AI applications using LangGraph and LLM technologies.
- Build agents capable of executing complex, multi-step business processes.
- Design stateful workflows incorporating reasoning, tool usage, validation, approvals, and exception handling.
- Develop single-agent and multi-agent solutions where appropriate.
- Translate business requirements into practical agentic AI architectures.
LLM Application Engineering
- Integrate LLMs into production applications.
- Develop prompt strategies, structured outputs, tool calling, and context-management approaches.
- Select appropriate models based on accuracy, capability, latency, security, and cost.
- Develop mechanisms to improve reliability and reduce hallucinations.
- Implement appropriate guardrails around AI-generated decisions and actions.
RAG and Enterprise Knowledge
- Design and implement Retrieval-Augmented Generation (RAG) solutions.
- Connect AI applications to enterprise documents, databases, APIs, and knowledge repositories.
- Develop retrieval and ranking strategies to provide agents with relevant context.
- Work with embeddings and vector databases.
- Implement data and context pipelines supporting AI agents.
Business Process Automation
- Analyse business processes and identify opportunities for agentic automation.
- Design AI workflows that combine LLM reasoning with deterministic business logic.
- Build agents capable of retrieving information, making decisions, invoking tools, and completing actions.
- Implement human-in-the-loop approval and escalation processes.
- Ensure automated actions are controlled, auditable, and reversible where appropriate.
Evaluation and Quality
- Develop evaluation frameworks for AI applications and agent workflows.
- Define metrics covering accuracy, task completion, reliability, latency, and cost.
- Build automated tests for prompts, agents, tools, and end-to-end workflows.
- Analyse failures and continuously improve agent behaviour.
- Use observability and evaluation data to optimise production systems.
Production Deployment
- Deploy and operate AI applications in cloud and enterprise environments.
- Implement monitoring, logging, tracing, and performance management.
- Design resilient workflows with retries, timeouts, fallbacks, and recovery mechanisms.
- Work with DevOps and platform teams to establish appropriate deployment and CI/CD practices.
Cross-Functional Collaboration
- Work closely with product managers, business analysts, software engineers, data scientists, architects, and business stakeholders.
- Communicate AI capabilities, limitations, risks, and implementation options.
- Help organisations identify realistic and valuable use cases for agentic AI.
Required Experience
- Commercial experience developing AI/LLM applications.
- Hands-on experience with LangGraph and agentic workflow development.
- Strong Python development experience.
- Experience deploying AI applications into production.
- Strong understanding of LLMs, RAG, tool calling, structured outputs, and prompt engineering.
- Experience integrating AI applications with APIs, databases, enterprise systems, or SaaS platforms.
- Experience with cloud platforms such as AWS, Azure, or GCP.
- Experience with AI evaluation, monitoring, and observability.
Desirable Skills
- LangChain / LangSmith
- Multi-agent systems
- AI workflow orchestration
- Vector databases
- Kubernetes
- Docker
- FastAPI
- Data pipelines
- MLOps
- AI security and governance
- Enterprise process automation
- Experience with financial services, healthcare, retail, manufacturing, or other complex enterprise environments
Originally posted on Himalayas