Open role · Remote, United States or Europe
Senior AI Engineer, LLM Applications and Agentic Systems.
Turn an agent idea into a system a client actually runs: existing models, real data, tool calling, guardrails, evaluation.
The mandate
From agent idea to governed production.
You do not train foundation models and this is not a research seat. You assemble existing models with the client's data, tools, rules and applications, and you make the result reliable enough to defend.
You build the category the house installs: the AI operating layer for the regulated enterprise. Audit-ready, regulator-ready, board-ready.
This is the strongest technical signal the house sends: operating systems in production, beyond audits, training and demonstrators.
The role, plainly
What you do. What we look for.
What you will do
- design agents that execute real business tasks, not demonstrations;
- build RAG systems on the client's document base;
- connect models to APIs, databases and internal software;
- implement tool calling and multi-step workflows;
- evaluate relevance, reliability and cost, with numbers;
- run guardrails, permissions and human validation mechanisms;
- deploy on Vertex AI, AWS Bedrock, Azure OpenAI or model APIs;
- own observability, testing and performance tracking.
What we look for
- an excellent software engineer in applied AI, not an academic data scientist;
- Python or TypeScript, model APIs, RAG, vector stores, evaluation;
- systems shipped to production and still defended, with their trade-offs;
- judgment on guardrails and human-in-the-loop design in regulated contexts;
- ten or more years in technology, three or more in applied AI;
- employment or senior independent mission: the model is aligned in writing at the Mutual fit stage.
The engagements you would carry
Real systems, in production, with your name on the decisions.
- An agent that reads regulatory documents and prepares a defensible business response.
- An internal assistant wired to the company's documentation and live data.
- An agent that searches, calls several APIs and prepares an action for human approval.
- An extraction and qualification system for case files, evaluated continuously.
- A copilot embedded in a Laravel, Drupal or Next.js business application.
Before you apply
Three screening questions.
Answer them in the application form. Numbers beat adjectives: stacks, metrics, failure modes.
- One agentic or RAG system you shipped to production: the stack, the guardrails, the evaluation, and what broke.
- How you measured reliability and cost on that system, with the numbers you tracked.
- Why governed AI for regulated clients, and why now?
Application
Show us one system you shipped and still defend.
A named partner reads every application within five business days. If we are not a fit, you get a clear reason in writing, not silence.