AI Engineer

Benford; Greater London; On-site; GBP 90,000 - 140,000; Full time

Benefits offered by this job

Competitive salary

Equity

On-site work in London

Job summary

Benford is building AuditOS to modernise audit work through AI-driven product engineering. You will own how AI performs real audit work from prototype to a client-ready platform, with the auditor signing off on the output and continual feedback weekly.

The role focuses on agentic document workflows, end-to-end problem solving, and a strong emphasis on traceability, control, and evidence around the models. This is a full-time, on-site London role with competitive salary and equity.

Qualifications

• Deep experience in LLM development or product engineering.
• Deliverable-focused—what you shipped matters more than tenure.
• Ability to operate across the stack when it counts.

Responsibilities

• Own agentic workflows over client accounting systems, from harness and evaluation to what an auditor actually sees.
• Solve end-to-end problems: pick it up, decide how to solve it and take it to market in the product.
• Own the control and evidence layer around the models, so inputs are logged and traced to a named auditor.
• Provide weekly feedback loop with the auditors who use what you build.

Skills

LLM development

Harness building

LLM product engineering

AI native development

Job description

You will own how AI does real audit work inside AuditOS, from the first prototype to something clients depend on. This is product engineering where the product is the audit itself.

Why Benford

Audit at the small and mid-market end has been abandoned while fees keep rising. The largest firms are stepping back, and every previous attempt to modernise the work aimed at the auditor's desk and left the firm untouched.

We are vertically integrated instead. We hold the licence, we build the platform and we employ the auditor who signs. AuditOS connects to the systems a company runs on, ERP, subledgers, invoices and banks, and carries the work from one year to the next so the second audit starts where the first finished. Routine work is done by systems, judgement stays with people.

What you'll own

• Agentic workflows over client accounting systems, from harness and evaluation to what an auditor actually sees.
• A problem end to end: you pick it up, decide how to solve it and take it to market in the product.
• The control and evidence layer around the models, so every input is logged, traced to its source and assigned to a named auditor.
• The feedback loop with the auditors who use what you build, every week.

Who we're looking for

• Depth in LLM development, harness building or LLM product engineering. Strong in one, able to operate across the stack when it counts.
• Evidence over seniority. What you have shipped matters more than years. Side projects count.
• You build with AI natively, not bolted on afterwards.
• You think about what does not exist yet, not about what is currently done.

Our stack

Fundamentals matter more than familiarity with any specific tool. If you are strong, you will pick it up.

• AI layer: agentic document workflows over client accounting systems, with every step logged and attributable to a named auditor. The hard part is not the model, it is control and evidence.
• Integrations: direct connections into ERPs, subledgers, banks and invoicing systems. Messy, real, and where most of the value is.

Practical

Full-time, on-site in London. Competitive salary and equity.