出海雷达

生成式AI管道工程师(技术负责人)

Generative AI Pipeline Engineer (Tech Lead)

AI开发工程职能支持限定地区(需当地身份)日间重叠约 2 小时,需偶尔早起或晚睡
公司CapsLock
薪资未公开
工作地点Poland
地域资格限定地区(需当地身份)
时区要求日间重叠约 2 小时,需偶尔早起或晚睡
用工类型Full Time
发布时间17 天前
数据来源Himalayas
前往 Himalayas 查看并投递 →
注意地域限制:该职位明确限定在 Poland 招聘。如果你是位于中国大陆的求职者,通常需要当地工作身份才能投递,或需与雇主确认是否接受独立合同(Contractor)形式合作。
作息提示:日间重叠约 2 小时,需偶尔早起或晚睡。

描述
CapsLock 是一家全球IT营销公司,为B2C客户提供独特且可扩展的客户获取解决方案。我们在多个行业的复杂销售挑战中提供解决方案,主要服务于北美大型合作伙伴。通过整合数字营销、IT、设计、AI驱动的分析、专有MarTech以及销售咨询方面的专业知识,我们提供强大的、数据驱动的客户获取解决方案。
在 CapsLock,人才、技术和前瞻性创新是我们一切工作的核心。我们的多元化、全球团队精通超过10种语言,注重开放协作和新想法的交流,不断突破界限以确保客户的成功。
职位概述
我们正在构建 CapsLock 的 AI 媒体制作平台,并正在招聘负责架构设计的技术负责人。
我们生成的是真实产品的视觉内容,而不是概念。每个素材都必须与客户的实际产品 1:1 匹配——包括硬件和几何结构。没有瑕疵,没有幻觉,没有“差不多”。这是一个控制和可重复性问题,而不是提示问题,需要工程师来解决。
这是一个从零开始的项目。你将是这个方向上的第一位工程师,直接与设计总监合作,作为创意团队和软件基础设施之间的技术接口。
你将负责的工作

  • 设计并运行我们的 AI 媒体制作平台:支持 ComfyUI 或类似引擎,GPU 环境,为非技术人员提供一键式用户界面;
  • 发布模块化、版本化、可重复使用的生成流程,基于节点或代码优先(例如 diffusers);
  • 开发图像和视频生成的一键预设和使用场景;
  • 端到端训练 LoRAs:从真实产品摄影到评估后的生产就绪模型;
  • 在每个流程中加入以控制为主的生成:ControlNet、IP-Adapter、修复、参考条件;
  • 将质量检查工程化到流程本身:在人工审核之前自动检查产品参考;
  • 构建我们的创意智能循环:将我们表现最佳的素材所具备的要素编码成可重复使用的预设和风格;
  • 与设计总监共同制定技术路线图;评估新模型和工具;通过赋能而非强制推动采用。

要求
必备条件
你不需要满足每一个要求。如果你在这些领域中的大多数表现出色,并能展示已交付的工作成果,我们会很高兴与你联系。

  • 已交付的生成流程被日常使用
查看英文原文

Description
CapsLock is a global IT marketing company that pioneers unique, scalable customer acquisition solutions for B2C clients. We solve complex sales challenges across a wide range of industries, primarily for large North American partners. By integrating our expertise in Digital Marketing, IT, Design, AI-driven analytics, proprietary MarTech, and sales consulting, we deliver powerful, data-informed customer acquisition solutions.
At CapsLock, people, technology, and forward-thinking innovation lie at the heart of everything we do. Our diverse, global team, fluent in over 10 languages, thrives on open collaboration and the exchange of new ideas, pushing boundaries to ensure our clients' success.
Role overview
We are building CapsLock's AI Media Production platform, and we're hiring the tech lead who will architect it.
We generate visuals of real products, not concepts. Every asset must match the client's actual product 1:1 - down to hardware and geometry. No artifacts, no hallucinations, no "close enough". This is a control and reproducibility problem, not a prompting problem, and it needs an engineer.
This is a greenfield build. You'll be the first engineer in this direction, working directly with the Design Director as the technical interface between creative teams and software infrastructure.
What you will work on

  • Architect and run our AI Media Production platform: serving ComfyUI or similar engines, GPU environments, one-click UIs for non-technical teams;
  • Ship modular, versioned, reproducible generative pipelines, node-based and/or code-first (e.g. diffusers);
  • Develop one-click presets and use cases for image and video generation;
  • Train LoRAs end-to-end: from real product photography to evaluated, production-ready models;
  • Build control-first generation into every pipeline: ControlNet, IP-Adapter, inpainting, reference conditioning;
  • Engineer QC into the pipeline itself: automated checks against product references, before a human ever reviews;
  • Build our creative intelligence loop: encode what makes our best-performing assets work into reusable presets and styles;
  • Own the technical roadmap with the Design Director; qualify new models and tools; drive adoption through enablement, not mandates.

Requirements
Your must haves
You don't need to meet every single requirement. If you're excellent in most of these areas and can show shipped work, we'd love to hear from you.

  • Shipped generative pipelines adopted and used daily by other people;
  • Production-grade workflows in ComfyUI or similar technologies, or equivalent code-first experience (e.g. diffusers);
  • Strong Python: automation, API integrations, custom nodes;
  • LoRA training end-to-end, from dataset curation to evaluation;
  • Control techniques across modalities (t2i, i2i, i2v, v2v): ControlNet, IP-Adapter, inpainting, reference conditioning;
  • Deploying pipelines as internal tools;
  • Enough visual literacy to set the quality bar (no design portfolio required);
  • Confident English (B2+).

Nice to have

  • 3D/render-hybrid workflows: Blender or CAD renders as geometric ground truth with generated environments;
  • LLM orchestration and agentic workflows: automated prompt generation, VLM-based QC;
  • Experience producing visuals for performance marketing.

What we offer

  • Remote Work - we offer a truly and fully remote environment. You choose where you are the most productive and comfortable to have an impact.
  • Paid vacations - generous paid vacation policy to ensure you have time to recharge
  • Unlimited Sick Days - we understand that being only human means getting sick or feeling under the weather from time to time, so we guarantee you time off as long as you need to recover and get back on your feet.
  • Ongoing Learning - people at CapsLock are deeply inquisitive and eager to learn new knowledge and skills, that's why we support and create learning opportunities like the Free Books program, workshops, conferences, and more.
  • Home Office - we will cover the equipment and furniture expenses to make sure you have the best work-from-home experience.
  • Physical Well–Being - We will cover costs of your medical insurance. Additionally we offer an allowance for a flexible fitness program.
  • Fun Stuff - there is never a shortage of fun stuff

Originally posted on Himalayas

本页面信息整理自 Himalayas,版权归原发布方所有。职位可能随时关闭,投递请以原始页面为准。 本站只做信息聚合展示,不参与招聘流程,也不向求职者收取任何费用。

该公司其他在招职位

资深 CRO 策略师

CapsLockUnited KingdomFull Time3 天前
市场运营限定地区(需当地身份)日间重叠仅 1 小时,需熬夜配合

产品经理

CapsLockCroatiaFull Time5 天前
开发工程职能支持未标注地域日间重叠约 2 小时,需偶尔早起或晚睡

← 返回全部职位