AI Engineer

AG Talent; United Kingdom; Remote; GBP 100,000 - 165,000; Full time

Job summary

AG Talent is seeking a senior AI Engineer to own the AI product area end-to-end. You will decide model strategies, build scalable systems, and drive live features from concept to production.

Work with the backend, data and DevOps teams, reporting to the CTO. You will choose providers, define the quality bar, and ensure robust guardrails before deployment. This is a fully remote role based in the UK.

Qualifications

• Strong backend engineering with Python and SQL proficiency.
• ≥18 months building with LLMs in production.
• Experience delivering features end-to-end in startups.
• Experience with multiple providers and model tradeoffs.

Responsibilities

• Build conversational retrieval over a large body of internal documentation to ground answers.
• Translate unstructured inputs (PDFs, images, spreadsheets) into structured data in the platform.
• Make live data queryable in plain English for non-SQL users, ensuring correctness.
• Choose and combine models across providers and explain tradeoffs.
• Design end-to-end retrieval: embeddings, chunking, vector search, document pipeline.
• Translate natural language into safe queries against live data and build guardrails.
• Integrate AI features into existing POS, dashboard and backend systems.

Skills

Python backend

LLMs in prod

RAG

Embeddings

Vector search

PostgreSQL

AWS

Tools

OpenAI

Claude

Gemini

Llama

Mistral

LangGraph

LangChain

Job description

Are you an AI Engineer with backend experience who has gone deep on LLMs and wants to own an entire AI product area rather than pick up tickets on someone else's?

A hospitality technology company whose platform runs live restaurant operations is building its first AI features, and needs one person to own all of them.

Fully remote, UK based.

3 stages:

• Conversation with the Founder,
• Technical Product Discussion with the Engineering Team,
• AI Assessment with an External Specialist including a technical test

Tech Stack: Python · OpenAI, Claude, Gemini, Llama, Mistral · RAG, embeddings and vector search · PostgreSQL · AWS

The platform is live, in production, and already carrying serious load. This is not a company bolting AI on to look modern. It is a company sitting on real operational data that AI can make genuinely more useful.

• Over 100 million transactions processed to date
• Around 500GB of live operational data across restaurant customers
• Series A closing, with revenue growing month on month
• An engineering team of eight: CTO, three backend, mobile, frontend, data and DevOps

The next step is turning that platform and that data into AI features restaurant operators actually use every day. That is where you come in.

You will own the AI product area end to end, as the first and only person in it.

The Role

You will work alongside the backend, data and DevOps engineers and report into the CTO, but the AI area is yours.

You decide which models suit which problem, you build the systems, you set the quality bar, and you take the features into production.

Nobody above you will dictate the provider, and nobody below you will build it for you.

What You Will Be Doing:

• Building conversational retrieval over a large body of internal documentation, so users get answers that are grounded and correct rather than plausible
• Turning messy customer-supplied files, PDFs, photographs, spreadsheets and whatever else arrives, into clean structured data inside the platform against a defined schema
• Making live operational data queryable in plain English by customers who do not write SQL, and being accountable for the answers being right
• Choosing and combining models across providers based on what each job needs, and being able to explain the tradeoffs
• Designing retrieval end to end: embeddings, chunking, vector search and the document pipeline behind it
• Translating natural language into safe, correct queries against live operational data
• Building the evaluation, guardrails and fallbacks that stop the system quietly getting things wrong in front of paying customers
• Integrating all of it into the existing point of sale, dashboard and backend systems

Experience We Are Looking For:

• Strong Python with a real backend engineering background behind it
• Around eighteen months or more building with LLMs properly, not occasional experiments alongside other work
• Features you have taken all the way into production, that real users depend on
• Hands on with more than one provider, and able to argue the case for one over another. OpenAI, Claude / Anthropic, Gemini, Llama, Mistral or similar
• RAG, embeddings, vector search and document retrieval
• Structured outputs, function and tool calling
• Extracting structured data from PDFs, images, CSVs and spreadsheets
• Strong SQL and PostgreSQL
• Exposure to evaluating output quality, hallucination, guardrails and fallback behaviour
• Comfortable deploying and running your own work on AWS or similar
• A startup or scaleup background where you owned something whole rather than a slice of it

Nice To Have

• Text to SQL or semantic layer experience
• LangGraph, LlamaIndex, LangChain or similar
• OCR, document parsing and multimodal models
• Vector databases and retrieval infrastructure at scale
• Evaluation datasets and automated regression testing
• Restaurant, hospitality, point of sale, payments or analytics experience
• Elixir or Phoenix, or experience working alongside an Elixir backend team

This is not the role for you if you want a large AI team around you, an established platform team to lean on, or a defined backlog handed over each sprint.

It is the role for you if you have been the person building the LLM features at your current company and you now want the whole area, the model decisions, and your name on it.

This is a senior role. Not a 9-5, not a 9-9-6. Judged on output.