AI Production Agent Engineer

Dex; Greater London; On-site; GBP 100,000 - 150,000; Full time

Job summary

Dex partners with leading financial institutions to build and license AI agents that automate critical customer lifecycle workflows. You will own the infrastructure that keeps production agents reliable, from evals and feedback loops to prompt engineering and fallbacks.

You’ll design human-in-the-loop decision logic and work across the full agentic and software stack, including Node and TypeScript, in a fast-growing London-based team.

Qualifications

• Production experience building and maintaining agentic or LLM infrastructure: evals, feedback loops, prompt engineering, fallbacks, and human-in-the-loop logic.
• Pragmatic ML foundation across supervised and self-supervised methods.
• Product-leaning and full-stack in mindset; comfortable contributing to a Node and TypeScript software stack.
• Strong system design capabilities; adaptable to TypeScript-based assessments.
• Thrives in a fast-paced startup environment, owning your lane while data strategy evolves.

Responsibilities

• Design and implement agent reliability infrastructure, including evals, feedback loops, prompt engineering, and fallbacks.
• Build and maintain human-in-the-loop decision logic for high-stakes agent actions.
• Take ownership of proprietary supervised and self-supervised ML models, applying recent academic work pragmatically.
• Contribute across the full agentic and software stack, including Node and TypeScript.
• Shape the evolving data strategy for production AI agents in a rapidly growing environment.

Skills

Agentic infra

ML fundamentals

Node & TypeScript

System design

Startup mindset

Tools

Node.js

TypeScript

Job description

Dex partners with leading financial institutions to build and license AI agents that automate critical customer lifecycle workflows. You will own the infrastructure that keeps production agents reliable, from evals and feedback loops to prompt engineering and fallbacks.

You’ll design human-in-the-loop decision logic and work across the full agentic and software stack, including Node and TypeScript, in a fast-growing London-based team.