Forward Deployed Engineer

What You'll Do

Own engagement delivery end to end. Scope with the department, design the solution, build it, deploy it, and agree the handover and ownership model - from prototype through to stable production. Engagements arrive as vague pain; you define the problem, not just the solution.

Engineer AI solutions properly. Pipelines, LLM API integration, evals, guardrails, monitoring, and cost and accuracy optimisation for the systems you build. Know when a step must be deterministic and when an LLM is the right tool.

Graduate tools into business systems. Take shared, team-load-bearing tools that people have built for themselves and rebuild them as owned business systems under a full SDLC where the value justifies it.

Deploy ML-built components into production. Serving, integration with the Moneybox platform, and everything surrounding the model, in partnership with Decisioning and Data Science teams (who own what happens inside the model) and our engineering squads.

Build reusable capability. Convert engagement learnings into shared tooling, templates, playbooks and self-serve workflows on the AI Platforms stack. Building out platform components including guardrails, sandboxing, workflows, and gateways.

Raise the bar. Work alongside embedded specialists during the team's ramp-up, absorbing and internalising their output so the capability stays with Moneybox.

In your first three months we expect your first departmental engagements to be selected on feasibility and delivered with measurable business value - time saved, cost avoided, risk removed - and at least one ML-built capability deployed to production with proper evals, monitoring and cost controls.

This role is explicitly not ML model training or data science, and it is not a chatbot-prompting generalist: this is production software engineering with AI at its core.

Who You Are

A production engineer with AI at the core. You have built and shipped LLM-powered systems that ran in production and can talk concretely about evals, failure modes, cost curves, and what you would do differently.
Comfortable in ambiguity. You can walk into a department with a vague problem and leave with a scoped, deliverable system, and you are managed by exception rather than by direction.
A strong partner to non-technical owners. Embedding, scoping and communicating with people who own a business process but not the technology is the job, not a distraction from it.
A platform thinker. You have turned one-off solutions into reusable tools or platforms before and you look for the pattern in every engagement.
Operationally minded. You care about monitoring, cost and reliability of what you ship, and about making a prototype into a system someone can rely on.
Safety-conscious by habit. PII handling, data-boundary discipline, prompt-injection awareness, human-in-the-loop design and graceful failure are how you build, not a checklist you apply afterwards.

Skills & Experience

Essential

5+ years of software engineering with meaningful production ownership.

Has built and shipped LLM-powered systems that ran in production.

Production-grade Python as your primary language.

LLM-powered systems in production: real-world experience working with models via APIs - harnesses, orchestration, structured output, tool use and agents, RAG where appropriate.

Evals and quality: designing evaluation sets, measuring accuracy, recall and precision for LLM steps, regression-testing prompts and workflows.

Safety and guardrails in practice: PII handling, data-boundary discipline, prompt-injection awareness, human-in-the-loop design.

Cost and performance optimisation: model selection, caching, batching, token economics, latency budgets.

Deployment and operations: CI/CD, containerisation, monitoring and alerting for AI workloads.

Data processing fundamentals: pipelines, transformation, validation, anomaly handling.

Customer- or stakeholder-facing delivery experience: consultancy, forward-deployed engineering, solutions engineering, or embedded or platform roles serving non-engineering users.

Desirable

Experience with Azure and/or .NET. We host on Azure and our core stack is .NET, so willingness to integrate with both is required; existing expertise is a bonus rather than a requirement.

Experience with agent hosting and sandboxing platforms, LLM or MCP gateways, or agentic workflow orchestration tooling.

Experience in financial services or another regulated environment.

Experience deploying models built by a data science team into customer-facing production systems.