维塔士
Senior AI Engineer
Senior AI Engineer
发布于 大约 13 小时前普通员工/个人贡献者
全职员工仅现场办公人工智能
人工智能
Agentic Ai
CI/CD
Comfyui
Generative Ai
Llm
Prompt Engineering
值不值得投
74综合评分
综合评分 74
前沿AI技术与游戏创意结合，技术成长空间大，但具体WLB和薪资福利需进一步确认。
从薪资福利、成长空间、工作节奏和岗位方向综合评估，方便横向比较。
薪资怎么样
40K
比人工智能中位数低
发布情况
新岗位 · 今天发布
公司情况
这家公司招聘很活跃
33 个在招
市场机会
选择面较广
上海 · 人工智能 · 约 110 个在招
职位详情
关于这个职位
这是维塔士公司的一个高级人工智能工程师岗位
您将作为核心贡献者，设计和构建用于视频游戏资产管线和制作工作流的智能体（Agentic）和生成式AI系统
您将深入参与内部智能体平台的技术架构，并与艺术家、技术艺术家等多方团队合作，将前沿AI能力转化为可用于生产的工具
最低要求
Foundation (must-have software-engineering baseline)
+ years of professional experience building production applications, with recent depth in AI/LLM-based or generative AI systems.
Strong proficiency in at least one of Python, TypeScript, or JavaScript - Python expertise is required for our stack, including FastAPI, Pydantic, SQLAlchemy, or equivalent.
Solid database skills across relational systems such as PostgreSQL and non-relational systems such as MongoDB or vector databases; familiar with caching and queues where applicable.
Working knowledge of secure API design, authentication, authorization, and input validation.
Comfortable with Git, GitLab CI/CD, and Docker/containers.
Proven testing mindset and experience with automated test suites such as pytest.
Agent loop / harness engineering
Demonstrated experience designing and operating agent loops in production - not just prompt-tuning a chatbot.
Deep, practical understanding of prompting, context engineering, guardrails, context poisoning or pollution, and the parameters that govern model behavior.
Hands-on experience integrating tools into agents through MCP, structured inputs/outputs, or sub-agent orchestration.
Experience with an agent development framework such as LangChain, LangGraph, Claude Agent SDK, Pydantic AI, or comparable.
Strong debugging instincts for non-deterministic systems and experience using practical logging, tracing, and evaluation techniques.
Creative AI
Hands-on experience building generative image, video, 3D, or multimodal workflows.
Experience with ComfyUI or a comparable workflow orchestration system.
Familiarity with model families such as Stable Diffusion, FLUX, Qwen, or comparable open-source or hosted generative models.
Experience with model customization or conditioning techniques such as LoRA, ControlNet, adapters, inpainting, pose, depth, reference images, or segmentation.
Ability to assess generated content for quality,
工作职责
Agent platform
Design and build key parts of our internal agent libraries - the core abstractions and developer ergonomics that let teams across the company build agents quickly and consistently.
Contribute to the architecture of our shared agent runtime and deployment patterns, helping teams run and maintain agents across production workflows.
Help define and evolve the agent loop / harness: prompt orchestration, tool invocation, sub-agent delegation, and recovery behavior.
Bring in reference patterns from the broader ecosystem, including open-source agent loops and harness projects, and adapt them to creative production use cases.
Agent loop & harness engineering
Drive prompting and context strategy at scale: system prompt design, guardrails, mitigation of context poisoning and pollution, and management of context windows and model parameters.
Design tool interfaces for agents: MCP servers, structured inputs/outputs for context, and sub-agent composition patterns.
Advocate for typed-agent frameworks, clear debugging patterns, and practical tracing for non-deterministic workflows.
Evaluate and integrate hosted and local LLM options where latency, cost, quality, or data-residency requirements demand it.
Creative AI workflows
Design, build, and maintain generative workflows for 2D assets, video, textures, concepts, and 3D-related content.
Evaluate and integrate image, video, multimodal, and 3D generation models for game-production use cases.
Develop and optimize node-based generation workflows using tools such as ComfyUI or comparable systems.
Customize generative models and workflows using techniques such as LoRA, ControlNet, adapters, inpainting, reference conditioning, depth, pose, or segmentation guidance.
Build workflows that improve controllability, style consistency, editability, and downstream production readiness.
Partner with artists and technical artists to translate creative requirements into repeatable AI-assisted workflows.
Creative tool and pipeline integration
Integrate agentic and generative AI systems with DCC tools and game-production environments such as Blender, Maya, Houdini, Unreal Engine, Unity, Substance, or comparable platforms.
Build plugins, scripts, services, APIs, or MCP-based integrations that connect AI systems with asset and content pipelines.
Connect generative workflows with asset-management, review, and publishing systems.
Work with art and production teams to ensure generated outputs meet project, format, and engine-ingestion requirements.
Agent memory
Design and build the key parts of the memory layer used across our agents: conversation history management, context chaining, and episodic memory.
Help define the boundary between short-term working context and long-term persistent memory.
Develop practical strategies for history compaction, retrieval, retention, and context reuse across production workflows.
Test- and eval-driven development
Build practical evaluation workflows for agentic and generative systems, including regression tests, representative prompts, workflow traces, and quality benchmarks.
Evaluate visual quality, prompt adherence, style consistency, temporal consistency, asset usability, and artist acceptance where applicable.
Build harnesses and CI checks that let us iterate on prompts, models, tools, and generation workflows with confidence.
Use measurable technical and production signals to guide improvements, including output acceptance, iteration speed, edit time, and workflow adoption.
Backend & platform foundations
Design and build scalable backend services and secure RESTful APIs in Python using FastAPI, with strong data modeling across relational and non-relational stores.
Apply appropriate authentication, authorization, input validation, and robust error handling for agent- and workflow-facing endpoints.
Implement caching, queues, retrieval systems, and vector storage where the workload requires it.
Build and optimize model-serving and inference workflows for hosted and locally deployed models.
Quality, delivery & collaboration
Drive performance tuning, code reviews, and technical documentation within your area of the AI platform.
Maintain CI/CD with Git/GitLab and Docker; ensure reproducible local-development and deployment pipelines.
Partner with artists, technical artists, UI/UX, production, engineering, and game-team stakeholders to translate workflows into AI-powered solutions.
Contribute to architectural decisions and share agentic- and creative-AI expertise with peers.
Work within agile methodologies.
AI 洞察
优缺点分析
优点
- 行业应用趣味性：将尖端AI技术应用于游戏这一充满创意的领域，工作成果直观可见
- 平台影响力：参与构建内部AI平台，其设计将影响公司内多个团队和项目
- 技术栈复杂且快速迭代：需要同时掌握后端工程、AI模型、游戏制作管线等多领域知识，且技术更新极快
- 跨职能协作要求高：需要频繁与艺术家、技术艺术家等创意团队沟通，将技术方案转化为创作友好的工具
- 对AI技术，尤其是LLM和生成式AI有深厚兴趣和实践，且不满足于调用API，追求深入系统设计的高级工程师
- 热爱游戏行业，希望用技术赋能创意制作流程，享受将前沿研究转化为实际生产力的过程
- 具备扎实的工程基础，能处理复杂系统，并乐于在一个需要紧密协作的环境中解决问题
缺点 / 挑战
- 技术前沿性：工作涉及最前沿的LLM、智能体架构和生成式AI技术，技术挑战和成长空间巨大
- 生产环境要求高：AI系统需要满足游戏制作对稳定性、质量、可控性的严格要求，调试非确定性系统有挑战
角色解读
- 技术专家路径：可深入AI架构设计，成为AI平台或创意AI领域的首席工程师或技术专家
- 技术管理路径：凭借丰富的项目经验和技术领导力，可晋升为AI团队或部门的技术经理
- 行业影响路径：在游戏AI或AIGC领域积累深厚影响力，参与行业标准制定或成为技术布道者
- 设计和构建用于游戏制作的智能体（Agentic）和生成式AI系统与工作流
- 负责内部AI平台的核心架构，包括Agent运行时、记忆层以及与DCC工具（如Maya，Blender）的集成
- 将前沿的LLM和扩散模型技术转化为艺术家和制作团队可用的生产工具
- 进行系统性能优化、代码评审和测试驱动开发，确保AI解决方案的稳定性和质量
- 精通Python后端开发（FastAPI, Pydantic），并有构建生产级应用（非聊天机器人）的经验
- 深度理解LLM应用开发，包括Prompt工程、上下文工程、Agent架构（LangChain/LangGraph等）
- 熟悉生成式AI工作流，尤其是基于ComfyUI的节点式工作流，并了解Stable Diffusion、LoRA、ControlNet等技术
- 掌握数据库、API安全、CI/CD和容器化等工程化技能
申请策略
- 研究维塔士的游戏项目和其在AIGC领域的公开信息或博客，了解其技术栈和业务重点
- 准备一个能展示你综合能力的个人作品或案例研究，最好涉及AI工具开发或创意工作流自动化
- 在面试中主动提问关于团队当前面临的最大技术挑战、内部AI平台的演进路线图，以及如何衡量AI工具在艺术家中的采纳度和效果，这能体现你的思考深度和岗位契合度
- 重点突出你在生产环境中构建、部署和维护LLM/生成式AI系统的具体经验，而非仅限于学术研究或原型
- 展示使用Python（FastAPI等）构建后端服务、API和数据库交互的完整项目经验
- 详细描述你使用或了解ComfyUI、Stable Diffusion、LangChain等工具或框架的具体项目，以及如何解决其中的技术难题
- 强调与工程、产品或设计团队协作，将AI技术集成到现有工作流中的成功案例
- 深入学习Agent架构：系统研究主流Agent框架（LangGraph, Claude Agent SDK）的设计思想，并动手实现一个简单的Agent应用
面试指南
- 请描述一个你设计或参与构建的、用于生产环境的AI Agent或生成式工作流系统
- 你遇到的最大技术挑战是什么？是如何解决的？
职位点评
薪资福利70
成长发展85
工作生活65
使命价值75
薪资福利
70中等
作为一家大型游戏技术公司的资深技术岗位，薪资应具有市场竞争力。但JD未明确提及薪资范围和具体福利细节。
薪资信号未披露（AI估算：30K-50K/月）
成长发展
85较高
该职位处于技术前沿，涉及LLM、智能体、生成式AI等尖端领域，且要求构建生产级系统，能提供极强的技术成长和项目历练。职责明确指向核心平台开发，具有明确的专业深耕路径。
技术前沿前沿/新兴技术
技术栈LLM、Agentic AI、Generative AI、FastAPI、LangChain、ComfyUI、Stable Diffusion、Prompt Engineering、MCP、Vector Databases
业务类型profit_center
工作生活
65中等
工作地点在中国（很可能为上海），但JD未明确说明办公模式和工作时间安排。大型项目制工作可能存在阶段性强度。办公地点预计在城市核心区或科技园。
工作模式未明确
办公地点科技园/产业园
加班情况未提及（无法判断）
使命价值
75中等
将AI技术应用于游戏制作，能为创意工作者赋能，提升创作效率与可能性，具有一定的创新价值和行业影响力。属于娱乐科技领域，使命主要体现在技术驱动的体验创新上。
行业发展稳定成熟行业
社会影响中性/一般
使命信号make games better and more immersive、pushing the boundaries of possibility、creating exciting new experiences
创新程度积极采用新技术
维塔士 的其他在招职位
相似职位推荐
Watch Jobs