You design, build and run agents that work inside our clients' operations: they read their information, act in their systems and stop wherever a person has to sign.
The role
- Build agents end to end on products already in production: finance, sales, projects and prospecting.
- Integrate models from different providers and make their behavior predictable in real cases, not in demos.
- Decide what is solved with an automation tool and what gets written in code, and defend the decision.
- Connect agents to what the client already uses: Google Workspace, WhatsApp, their usual systems, their spreadsheets.
Your first 90 days
- Day 30
You learn the platform and the six agents already in operation. You ship your first improvement to production, with its test.
- Day 60
You take ownership of a whole agent: its tools, its limits and its log.
- Day 90
You take an agent into a client's operation and measure what changed in their process.
Stack and tools
Claude APITypeScriptNext.jsSupabase / PostgresPythonClaude CodeWebhooks and third-party APIs
And what our clients already use
What we look for
Must have
- You have built something with language models that another person uses, even if it is a side project.
- You work with incomplete documentation and get there by reading, testing and measuring.
- You can tell an agent that seems to work from one that still works on day 30.
- You write code someone else can read and maintain.
Nice to have
- Experience with RAG, semantic search or model evaluation.
- Having connected company systems to each other.
- Product judgment: you know when to say something should not be automated.
What working here is like
What you build goes into products that operate real organizations. You work directly with the founders, with no layers between your code and production, under the same rule we give the intelligence: a person signs what is critical.