Full-stack AI engineer: coding agents, GEO platform

AISEO builds the platform that gets businesses mentioned by AI. When someone asks ChatGPT, Claude, or Perplexity for a recommendation, we make sure our customers are in the answer. This is Generative Engine Optimization (GEO), and we think it replaces a large part of what SEO used to be. We are a small team and we build the way the best small teams build now: engineers driving fleets of coding agents. We are hiring one more person who works this way natively. What this role is: You build the agents inside our GEO platform. Crawlers that check how LLMs talk about a brand, pipelines that generate and place content answer engines actually cite, monitors that track mentions across ChatGPT, Claude, Perplexity, Gemini, and AI Overviews, and tooling that makes a business legible to LLMs (think llms.txt, structured docs, machine-readable product data). You own these systems end to end: design, build, ship, measure. You also build with agents, not just for them. We expect you to run Claude Code or Codex as your primary way of producing software, and to be good at loop engineering: turning a goal into a prompt plus a verifiable completion condition, then letting agents iterate overnight until it holds. Backlog burndowns, fix-until-green loops, eval-driven content pipelines. If you have never left an agent running while you slept, this role will be a stretch. What we look for: - You use coding agents as your default. You decompose work, write prompts that hold up over long runs, manage context windows, and run several agents in parallel without losing the thread. - You have built agent loops with real completion criteria, not just chat sessions. You know why "keep going until the tests pass" needs guardrails and what those guardrails are. - You have shipped LLM-powered products: prompting, evals, retrieval, structured output, cost and latency tradeoffs, multi-provider routing (OpenAI, Anthropic, open models). - You are a solid full-stack engineer underneath it all: TypeScript, React/Next.js, Node.js, and Python, with production APIs, data models, and deploys (GCP preferred) behind you. - You read every diff an agent produces and catch its shortcuts. Agent fluency without engineering judgment is how bad code ships fast. Nice to have: - You understand how AI is changing search: AI Overviews, answer engines, citation behavior, semantic search, vector databases. - You have experimented with MCP servers, agent skills, or tool-use APIs. - You have taken a side project from idea to launch on your own. - You have opinions about how businesses should present themselves to LLMs, and can defend them. What you will work on: - Agents that audit and improve a brand's presence in LLM answers. - Content and data pipelines optimized to get cited by answer engines. - Tooling that makes customers' products discoverable and citable by LLMs. - The evals and monitoring that prove all of the above actually moves mentions. How to apply: Send two things. 1. Three pieces of evidence of exceptional ability. Anything that proves you operate at a level most engineers do not: products you shipped solo, agents you built, loops that ran unattended and produced real results, open-source work people use, numbers you moved. Links and artifacts beat descriptions. Pick your best three, not your most recent three. 2. Your understanding of GEO (max 300 words). How do LLMs decide which businesses to mention in an answer? What actually moves citations in ChatGPT, Claude, or Perplexity, and what is noise? What agent would you build first to get a brand mentioned more, and how would you know it worked? We want to see you have thought about this, not just read a blog post about it. We read for evidence and judgment. A strong submission from someone with three years of experience beats a generic one from someone with ten.