你将参与的事：

我们正在把大语言模型的推理与执行能力嵌入真实的业务流，让影像创作、内容生产等场景能被 AI Agent 更自主地驱动。你将为这套 Agent 系统设计灵活且高可用的架构，让它不止跑通 demo，而是在千万用户量级下稳定运转。

核心工作：

· 主导 AI Agent 应用的整体架构设计与核心模块实现，让 Agent 具备可靠的规划、工具调用与自主执行能力；

· 持续打磨系统性能与稳定性，在高并发、高可用的目标下落地具体优化方案；

· 与产品、算法和前端同学紧密协作，把用户场景从想法变成链路完整、体验细腻的 AI 交互体验。

我们希望你具备：

· 3 年以上后端开发经验，有 AI Agent 项目的实际交付经历；

· 精通 Python，熟悉主流 Web 服务框架，对高并发、高可用系统的设计和实现有扎实体感；

· 熟悉主流的 Agent 框架，对 RAG、Function Calling、MCP 协议等相关技术有深入理解和实践经验；

· 清楚 LLM 的基本原理，能熟练使用 Prompt Engineering 来解决工程问题，而不只是调用 API；

· 习惯在标准的团队协作流程里工作，能务实推动问题闭环，并持续更新自己的技术认知。

加分项：

· 有 Dify、ComfyUI 或其他 AI 工作流编排框架的实战经验；

· 对影像处理、多模态内容生成有天然的好奇心，喜欢用技术创造好玩或好用的东西。

What You’ll Be Part Of:

Meitu builds AI-powered imaging tools used by millions every day. We’re looking for an engineer who treats AI Agents not as black boxes, but as systems to design, optimize, and push further. You’ll join the team that transforms the latest reasoning and tool-use capabilities into features that actually ship — and feel magical to use.

How You’ll Contribute:

· Architect and build the core of AI Agent applications — from requests flowing through function calls to the final output users see, you own the system design, module implementation, and performance tuning;

· Prototype agents that connect LLMs to real-world tools using RAG, Function Calling, and protocol-driven interactions, then harden them into production-ready services;

· Collaborate with product, algorithm, and frontend engineers to turn business requirements into technical solutions that are both robust and delightful to use;

· Optimize the end-to-end pipeline — reduce latency under load, improve reliability of tool execution, and ensure the agent behaves predictably even at scale;

· Stay hands-on with the agent ecosystem: experiment with new frameworks, contribute to evaluation benchmarks, and share what you learn with the team.

What We Hope You Have:

· 3+ years of backend development experience, including projects where you shipped a real AI Agent or LLM-powered application into production;

· Deep fluency in Python, web frameworks (any common stack), and the patterns needed to build high-concurrency, high-availability systems;

· Strong working knowledge of mainstream agent architectures — you’re familiar with RAG, Function Calling, MCP, and can discuss the trade-offs of different orchestration choices;

· Solid grasp of LLM fundamentals; you write and refine prompts with intention, not guesswork;

· You follow standard engineering practices (version control, code review, CI/CD) and communicate clearly with teammates — you’re the kind of engineer others enjoy working with.

Nice to Have:

· Experience with AI workflow or low-code agent frameworks like Dify, ComfyUI, or similar;

· Contributions to open-source projects in the LLM or agent space;

· Comfort with infrastructure tooling (containerization, observability) in a cloud environment.