> Markdown version of [/jobs/ext/4018901-ai-engineer](https://www.wearedevelopers.com/jobs/ext/4018901-ai-engineer). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# AI Engineer

- **Company:** Trivium
- **Location:** UK (Remote available)
- **Experience:** Expert
- **Salary:** £39,986.0
- **Contract:** Contract
- **Skills:** Application Programming Interfaces (APIs), Artificial Intelligence, Cursor, Software Debugging, Python (Programming Language), PostgreSQL, MongoDB, TypeScript, Website Builder, Large Language Models, Claude Code, AI Coding Agents, Langfuse — LLM Observability and Analytics Platform, Agentic-AI, Virtual Agents
- **Published:** October 30, 2026
- **Apply:** https://www.adzuna.co.uk/jobs/details/5946151027

## About the Role

* 7+ years building and operating production systems, with real on-call and incident ownership.
 * Hands-on LLM applications in production: tool calling, prompt and context debugging, evals (Langfuse or similar).
 * Daily, disciplined use of AI coding tools (Claude Code, Codex, Cursor), including running several agents in parallel.
 * Strong Python and TypeScript; event-driven services; MongoDB or Postgres.
 * Experience with messy third-party integrations (scheduling systems, telephony, marketing and ad platforms).
 * Ability to work U.S. Eastern Time business hours.
 * Based in Canada, Argentina, Brazil or the United Kingdom, and able to work as an independent contractor.

## Description

An engineer who takes full ownership of production and, when something breaks, knows exactly what went wrong and where, usually before the practice notices.

 * You run multiple AI agents, well. You direct several AI coding agents in parallel, review what they produce critically, and keep a codebase clean while moving fast. You also own our production agents (voice, reviews, reporting) as systems, not demos.
 * You know where AI helps and where it doesn't. You decide deliberately when to delegate to an AI agent and when to write the code yourself: tricky concurrency, scheduling logic, security, anything where being subtly wrong is expensive.
 * You debug fast and systematically. You form a hypothesis, check the logs and traces, and find the root cause, whether it's in our code, a vendor API, or the model's prompt and context.
 * You build the monitoring that's missing. If you can't see it, you instrument it: tracing across every call and agent action, alerts that fire on real failures, dashboards people actually use.
 * You can pinpoint where an LLM went wrong: which prompt section, which tool call, which piece of context. Then you prove the fix with evaluations and replays of real failures.
 * You own outcomes. You don't wait for tickets; you find what's broken, fix it, and make sure it stays fixed.

WHAT YOU'LL WORK ON

 * Production reliability and incident response across our AI agents and integrations.
 * Debugging and evaluation tooling for LLM agents (prompt, context and tool-call tracing; regression evals; failure replay).
 * A real-time availability viewer and configurator: every office schedules differently, so make it visible and configurable exactly what the voice agent will offer, and why.
 * Helping finish and harden the Marketing Report and the Website Builder.