Acorn Insurance is hiring an AI Engineer for our Fraud AI Team

This is a focused, hands-on role with one clear opening priority: take ownership of our Neo4J-based fraud detection solution. After that, there is a meaningful backlog of Fraud AI work waiting - detection workflows, casework support, signal generation - and the engineer who takes this role will be the one shaping how we deliver it.

Acorn is FCA-regulated, with Consumer Duty and UK GDPR shaping how AI gets built and run. Fraud work in particular needs to be auditable and defensible - that is a design constraint, not an afterthought.

Fraud is one of the highest-stakes places AI gets deployed in insurance. The work you do here directly affects detection rates, investigator productivity and customer outcomes.

Job Title: AI Engineer – Fraud

Salary: Up to £45,000 depending on experience

Working hours: 37.5 hours per week, Monday to Friday 9am to 5:30pm

Location: Liverpool City Centre on a Hybrid working basis

What you will be doing:

• Take ownership of the Neo4J network graph solution for Fraud: schema, ingestion, queries, performance, evolution.
• Work through the Fraud AI backlog - detection workflows, casework support, fraud signal generation, investigator-facing tooling.
• Build and ship in n8n and AI Foundry / Requesty, integrating Neo4J outputs into investigator workflows.
• Partner with the existing Fraud AI Engineer on priorities, design and delivery; share load and context.
• Stakeholder management with Fraud ops and the Fraud SLT - translate operational concerns into deliverable work.
• Make sure outputs are auditable and defensible - documentation, logging, evaluation and human-in-the-loop where the risk profile demands it.
• Lean on the Central / Group team for platform support, observability standards and burst capacity when needed.
• Work within a regulated AI governance model: DPIAs, model documentation, observability, human-in-the-loop where appropriate.

What we are looking for:

• Solid engineering background with production experience shipping software or AI / ML-backed systems. You've built things people rely on, not just prototypes.
• Strong working knowledge of graph databases - Neo4J specifically, or comparable (TigerGraph, Amazon Neptune). Cypher fluency is a real plus.
• Databricks experience is a strong plus - it underpins our wider data platform and Fraud AI work increasingly touches it.
• Comfort with Python, REST / webhook integration, JSON, and at least one workflow / orchestration tool (n8n preferred given our stack).
• Working comfort with YAML - it shows up everywhere in configuration, CI and infrastructure-as-code. Easy to pick up; expected as table stakes.
• Working knowledge of LLM application patterns: RAG, structured extraction, evaluation, agentic workflows - enough to use them where they're the right tool.
• Domain exposure to insurance fraud, financial crime, or a comparable detection / investigation function is a strong advantage.
• Awareness of the constraints of regulated environments: auditability, defensibility, data residency, model risk - design around them, don't work against them.
• Comfort working in a small team where you own outcomes end-to-end - hands-on delivery, not just direction.
• Cloud familiarity (Azure preferred) without needing to be a platform engineer - that work sits with the Central / Group team.
• Formal AI/ML credentials are not a hard requirement. Demonstrable delivery, sound judgement and the ability to ship into a regulated environment matter more.