Applied AI Engineer

About the role

We help large enterprises cut energy costs and valorize flexibility in the age of AI. We're hiring an Applied AI Engineer to own the foundation under our agentic features: the architecture, the harnesses, the evals, the tool surface, the model choices and the guardrails that let AI workflows run reliably for large enterprise customers. Location: Ghent, Belgium (hybrid).

Responsibilities

• Define the agentic architecture: how an agent is defined, how tools are exposed to it, how context gets assembled, how a run is traced, and how state and long‑running work are handled.
• Build evaluation harnesses and datasets to decide what “good” means for a feature whose output differs on every run, and catch regressions before a customer does.
• Design the tool surface for models (MCP, internal management server, customer‑facing analytics server), deciding which workflow steps should stay deterministic code.
• Manage multi‑model usage: selecting which model runs each step, enabling swaps based on evaluation evidence rather than release notes.
• Apply cost, latency and reliability as product constraints: budget per feature, caching, retries, fallbacks, and monitoring that flags when a workflow becomes slower or more expensive than its value.
• Implement guardrails our customers can accept: permission‑aware, tenant‑isolated, auditable, with human‑in‑the‑loop wherever a wrong number could turn into a claim against a supplier.
• Develop workflows on top of the foundation, e.g., contract and hedge position ingestion: parse documents, propose mapping, let a human confirm.

Requirements

• Built agentic systems that have survived contact with paying customers: tool design, context engineering, multi‑step and long‑running flows, structured output, retrieval where it adds value, and judgment to keep steps deterministic when appropriate.
• Operated agentic systems after launch: managing evals, regressions, and surprises from real users.
• Created eval harnesses for non‑deterministic systems, defined “good” before shipping, and used them to catch regressions.
• Implemented guardrails around business‑critical data: permission‑aware, tenant‑isolated, auditable, with human‑in‑the‑loop where a wrong answer costs money.
• Designed tool surfaces for models rather than people: hands‑on with MCP and agent SDKs, understanding that exposing a good tool is a design problem.
• Strong software engineering in Python: writing maintainable code, with opinions on tests, CI, and rollback.
• Pluses: document understanding at scale (extracting from messy real‑world PDFs), comfort with TypeScript and React for feature UX, streaming/long‑running task UX, SQL and time‑series data at volume (PostgreSQL/TimescaleDB or similar), knowledge of energy systems (electricity markets, contracts, tariffs, grid fees, load profiles), prior experience in a fast‑growing startup or scale‑up.