Forward Deployed Engineer, Agentic AI

present of AI; Regno Unito; In loco; GBP 65.000 - 90.000; Tempo pieno

Descrizione del lavoro

PolyAI in the United Kingdom is seeking a mid-level engineer who has built and shipped LLM agents to real users. You will own customer agents from scoping through production, handle tool calls, and debug live calls across speech recognition, models, prompts, and telephony integrations.

You should write solid Python, integrate enterprise systems and telephony, run eval sets and A/B tests, and communicate trade-offs clearly to engineers and executives. This is hands-on, production-focused work.

Competenze

• Shipping LLM agents in production.
• Strong Python skills.
• Experience integrating with telephony systems.
• Ability to run eval sets and online experiments.

Mansioni

• Own customer agents from scoping to production.
• Debug live calls across speech, model, prompts, and lines.
• Explain trade-offs clearly to engineers and executives.
• Implement tool calls and guardrails for agent behavior.

Conoscenze

LLM agents

Python

Telephony integration

Production debugging

A/B testing

Strumenti

SIP

Telephony platforms

Descrizione del lavoro

Customer & Solutions · United Kingdom · Mid level

This role asks for an engineer who has already built and shipped LLM agents to real users, including tool use and evals, and who can own a customer's agents from first scoping through production and beyond. A strong candidate writes solid Python, integrates messy enterprise systems and telephony, and debugs live calls to find whether the fault sits in speech recognition, the model, a prompt, an integration or the phone line. The person also needs to run agents like a scientist, using eval sets and A/B tests, and to explain trade-offs plainly to both engineers and executives.

How it fits the AI era

The work sits at the hard end of applied AI: real-time voice agents on live phone lines for large brands, where callers notice every mistake. AI handles much of the conversation, but the engineer still designs behavior, guardrails and tool calls, judges what went wrong on a bad call, and decides what ships based on data. The human parts are ownership, judgment and clear communication with customers.

How it builds AI experience

Someone here would gain hands-on experience shipping LLM agents in production, building eval sets from real calls, and running online experiments. They would also learn voice and telephony work across SIP and contact-centre platforms, and get close to Raven, the company's own dialog model, working with the team that trains it. Building that experience means showing a shipped agent, strong Python, and comfort with production integrations.

Worth showing

• Shipped LLM agents
• Tool use and evals
• Strong Python
• Live call debugging
• Customer-facing ownership

This is our own read, written from the employer's posting and the company's Spotlight. The employer's posting is the source of truth for the role.

About PolyAI

PolyAI builds dialog agents that handle critical enterprise customer conversations and resolve them without escalation, per the company.