高级AI构建师
Senior AI Builder
我们是一家国际金融科技集团，在核心市场运营多家做市商公司，总部位于迪拜并正在国际化扩展。业务发展迅速，AI 是我们运营和扩展的核心。
高级 AI 构建者（代理与自动化）
全职 · 远程 · 金融科技 / 做市交易 · 隶属于迪拜的集团
高杠杆人选：一位优秀的构建者能实现整个团队的产出。你负责整个业务中的 AI —— 支持、内容、获取、运营 —— 并将模糊的业务问题转化为可部署并运行在生产环境中的系统。我们对你的期望很高，并会相应地提供资源。
你构建，我们部署。你直接向创始人汇报。
你将负责的内容
支持，接近自主
一个在 Intercom 上运行的代理，能够读取完整的知识库并端到端解决大部分票务，覆盖我们市场中的各种语言，对剩余问题进行升级，并已收集上下文信息。
内容引擎
生成、测试和迭代广告创意、落地页文案、社交媒体和 Reddit 帖子以及生成视频的流程，相当于一个完整内容团队的产出。
数据驱动的获取
跨 WooCommerce 和 Meta 的支出代理，实时跟踪每个国家的 CPA 和 ROAS，标记需要削减和扩大的部分，并每天早上提供决策简报。
持续的竞争情报
监控其他做市商 —— 定价、规则、促销、 payout 变化 —— 并提供每日摘要。
定制内部软件
使用 Claude Code，将真实工具部署到生产环境 —— 插件、内部应用、爬虫、仪表盘。不需要开发团队。
多代理编排
能够规划、委派给子代理、自我纠正，并具备记忆和自我提升能力，全天候运行的系统。
标准（我们根据实际生产成果招聘，而不是演示）
· 每天高级使用 Claude：Claude Code、Claude Cowork、API。与 ChatGPT 对话不是一项技能。
· 你已经构建、部署并运行了 AI 代理在生产环境中，而不是一次性演示：OpenClaw、Hermes Agent、自定义构建或其他类似项目。包括设置、记忆、监控、成本控制、可用性。
· 你交付真实的软件。你能阅读、编写和调试 Python 和 JavaScript，并将一个想法转化为可工作的工具，无需等待他人。
· 你了解当前的模型格局以及何时使用什么：Claude Opus 4.8 / Fable 5，GPT-5.5，Gemini 3，开放权重模型如 Qwen。你每周关注发布。
· 端到端自动化：n8n、Make 或代码，APIs，webhooks，爬虫，MCP。
· 使用当前顶级模型进行 AI 图像和视频生成：Nano Banana Pro、FLUX.2、GPT Image
查看英文原文
We are an international fintech group operating several prop trading firms across our core markets, headquartered in Dubai and scaling internationally. The business is growing fast, and AI is central to how we operate and scale.
Senior AI Builder (Agents & Automation)
Full-time · Remote · Fintech / Prop Trading · Dubai-based group
A high-leverage hire: one strong builder delivering the output of an entire team. You own AI across the business — support, content, acquisition, operations — and turn vague business problems into systems that ship and run in production. We expect a lot and resource you accordingly.
You build, we deploy. You report directly to the founder.
What you will own
Support, near-autonomous
An agent on Intercom that reads the full knowledge base and resolves the bulk of tickets end to end, across our markets' languages, escalating the rest with context already gathered.
A content engine
Pipelines that produce, test and iterate ad creatives, landing copy, social and Reddit posts, and generated video, at the output of a full content team.
Data-driven acquisition
Agents that cross WooCommerce and Meta spend, track CPA and ROAS per country in real time, flag what to cut and what to scale, and deliver a decision brief every morning.
Permanent competitive intelligence
Monitoring of other prop firms — pricing, rules, promos, payout changes — with a daily digest.
Custom internal software
With Claude Code, real tools shipped to production — plugins, internal apps, scrapers, dashboards. No dev team required.
Multi-agent orchestration
Systems that plan, delegate to sub-agents, self-correct, and run 24/7 with memory and self-improving skills.
The bar (we hire on proof of production, not demos)
· Daily, advanced use of Claude: Claude Code, Claude Cowork, the API. Chatting with ChatGPT is not a skill.
· You have built, deployed and run AI agents in production, not one-off demos: OpenClaw, Hermes Agent, custom builds or equivalent. Setup, memory, monitoring, cost control, uptime.
· You ship real software. You read, write and debug Python and JavaScript and turn an idea into a working tool without waiting on anyone.
· You know the current model landscape and when to use what: Claude Opus 4.8 / Fable 5, GPT-5.5, Gemini 3, open-weight models like Qwen. You follow releases weekly.
· End-to-end automations: n8n, Make or code, APIs, webhooks, scraping, MCP.
· AI image and video generation with the current top models: Nano Banana Pro, FLUX.2, GPT Image 2, Ideogram (image); Veo 3.1, Kling 3.0, Seedance 2.0, Runway (video).
· You manage what breaks: hallucinations, runaway API costs, edge cases, reliability over time.
· Fluent English.
· A portfolio of real, shipped, running systems. No portfolio, no reply.
What we do not care about
· Diplomas, AI certifications, prompt engineering courses.
· We only look at what you built and what it survived in production.
What you get
· Full tools budget: Claude Max, API credits, whatever your stack needs.
· Real ownership and a direct line to the founder. You decide how things get built.
This role is not for you if
· You have only "used ChatGPT a lot."
· Your war stories are vague.
· Your projects look good in a demo but have never run in production.
· You wait to be told exactly what to do.
· You want a quiet job.
How to apply
To confirm you read this post, start your message with the word AGENT. Applications without it are not read.
Send:
· 2 or 3 systems you built and ran in production: the problem, the stack, the result, and what broke and how you handled it. Links, demos or repos required.
· In a few lines: with your first 30 days here, what AI system would you build first to create measurable business impact in a prop firm?
Process
· Shortlist
· Screen-share call
· Offer