Software Engineer

DMCG Global; Greater London; On-site; GBP 100,000 - 140,000; Part time

Job summary

DMCG Global is seeking an experienced Agentic Software Engineer (Claude) for a contract role. The position is predominantly remote, with travel to London about once a month, and starts with an initial three-month term likely to be extended.

You will own end-to-end delivery of agentic AI solutions, using Claude within modern coding environments to translate business requirements into concrete outputs. The role requires building and testing agents, identifying failure modes, implementing

Qualifications

• End-to-end agentic AI delivery experience.
• Strong Claude experience within a modern agentic coding environment.
• Experience in scaling startup environments.
• Sound judgement around AI reliability, risk and failure modes.
• Structured testing, verification and guardrails approach.
• Broad technical skills across the software delivery stack.
• Ability to work independently with ownership.
• Strong stakeholder and requirements gathering skills.
• Ability to translate unclear business problems into technical deliverables.
• HR/payroll experience is a plus.

Responsibilities

• Build, extend and maintain AI agents and tools.
• Use Claude and modern agentic coding environments to direct delivery.
• Design agents that convert business requirements into outputs.
• Identify errors, hallucinations and AI content weaknesses.
• Test failure modes and validate outputs reliability.
• Implement fail-closed guardrails.
• Establish verification and QA processes.
• Run requirements-gathering sessions with stakeholders.
• Translate requirements into backlog items and acceptance criteria.
• Prioritize work and create delivery plans.
• Communicate progress, risks and decisions.
• Own delivery from idea to testing and production.

Skills

Agentic AI delivery

Claude experience

Scaling startup

AI reliability

Guardrails testing

Requirements gathering

Backlog prioritization

Stakeholder communication

Independent ownership

End-to-end delivery

Full lifecycle

HR/payroll helpful

Tools

Claude

Job description

Agentic Software Engineer (Claude) | Contract

Predominantly remote, with travel to London around once a month

Initial three-month contract, with a likely three-month extension

I have a genuinely exciting opportunity with an excellent scaling company that is investing heavily in its internal AI capability.

This is a hands-on role for someone who has already designed, built and delivered agentic AI solutions end to end. It will not suit someone whose experience is limited to using Claude as a general coding assistant.

What you’ll be doing:

• Building, extending and maintaining AI agents and tools
• Using Claude and modern agentic coding environments to direct delivery
• Designing agents that turn business requirements into completed outputs
• Identifying errors, hallucinations and weaknesses in AI-generated content
• Testing failure modes and challenging the reliability of outputs
• Implementing fail-closed guardrails
• Establishing verification and quality assurance processes
• Running requirements-gathering sessions with business stakeholders
• Translating requirements into backlog items, user stories and acceptance criteria
• Prioritising work and creating clear delivery plans
• Communicating progress, risks and technical decisions
• Owning delivery from the initial idea through to testing and production

What you’ll need:

• Demonstrable end-to-end agentic AI delivery experience
• Strong experience using Claude within a modern agentic coding environment
• Experience working within a scaling start-up environment
• Sound judgement around AI reliability, risk and failure modes
• A structured approach to testing, verification and guardrails
• Technical breadth across the software delivery stack
• The ability to work independently and take genuine ownership
• Strong stakeholder and requirements-gathering skills
• The ability to turn unclear business problems into practical technical deliverables

Experience within HR or payroll would be useful, but it isn’t essential.

The client needs someone who can do more than build a prototype. You must be able to take an agentic AI product through the full lifecycle and demonstrate how you ensured its outputs were accurate, reliable and fit for purpose.