AI Engineer

Prism Digital; Greater London; Hybrid; GBP 827,000 - 1,240,000; Part time

Job summary

Prism Digital on behalf of a prestigious financial research client in London seeks an AI Engineer to design and deliver agentic AI for risk analysis. You will be onsite two days a week, with remote collaboration through a consultancy, and start immediately to build the first wave of an important new capability.

You will translate risk experts' requirements into robust architectures, own end-to-end design, and work with Claude Code, Git, vector data sources and live risk data to deliver automated

Qualifications

• Context engineering: prompt construction, context window management, vector retrieval, information routing.
• Agentic orchestration frameworks: LangGraph, Pydantic AI, AutoGen or similar.
• Multi-agent systems, tool-use mechanisms and multi-step workflows.
• Python – very strong experience.
• AI evaluation and observability: agent accuracy, LLM output quality, regression detection, tracing.
• Translating requirements directly from non-technical subject matter experts.
• Senior-level ownership of architecture and solution design.

Responsibilities

• Translate risk requirements from non-technical stakeholders into AI agent designs.
• Own architecture and deliver production-ready agentic software.
• Collaborate with risk, cyber and safety stakeholders in London on-site.
• Integrate with Claude Code, Git, and data sources to deliver automated risk assessments.

Skills

Context engineering

Orchestration frameworks

Multi-agent systems

Python

AI evaluation

Requirements translation

Architectural ownership

Tools

Claude Code

Git

Jira REST APIs

Confluence APIs

Vector databases

Job description

AI Engineer | Python, Agentic AI | Finance House – Risk Team

• IR35: Very likely to be inside IR35
• Duration: 6 months -12 months, very likely to extend
• Location: London, two days a week on client site
• Start: Immediate, so short-notice availability matters

You will be contracted via by a global digital engineering consultancy (prestigious one!) onsite at one of their clients, a financial research and risk organisation in London, to build the AI agents their risk function will run on.

The risk is manually performing its tasks with humans, reading across twenty or thirty separate data sources and forming a judgement. You will build the agentic systems that take those decisions on: sitting with the people who hold the risk expertise, understanding how they reason, then turning that into software which assesses, triangulates and reports on their behalf. The client is tight-lipped about what it builds, so expect to learn most of the detail once you are through the door.

The risk function has no engineering capability at all right now. You are the first hire in, proving out something both sides expect to grow into a large programme of work. The skillset is scarce inside the consultancy too, so there is a route into their other accounts from here.

This suits an engineer who wants to be in the room with people. Most days you will be in front of non-technical operational risk, cyber and safety stakeholders, drawing the requirement out of them before you write a line of code, then going away and building it properly. You need the depth to own the architecture and the presence to be the AI expert in front of a client.

What They're Looking For (Non-Negotiables)

• Context engineering: prompt construction, context window management, vector retrieval, information routing
• Agentic orchestration frameworks: LangGraph, Pydantic AI, AutoGen or similar
• Multi-agent systems, tool-use mechanisms and multi-step workflows
• Python – very strong experience
• AI evaluation and observability: agent accuracy, LLM output quality, regression detection, tracing
• Translating requirements directly from non-technical subject matter experts
• Seniority to own architecture and solution design, not only implementation

Experience working inside a risk function, or in the risk industry generally. They are not asking for a risk expert, they have those already. But an engineer who already speaks the language will get productive far faster here, and it counts for a lot. If you have it, lead with it.

What You'll Work With

• Claude Code, which the client uses heavily and treats as the default way of writing software
• Git, Jira REST APIs and Confluence APIs
• Vector databases and retrieval layers
• 10-30 upstream data sources feeding a single automated assessment core
• Live risk registers and executive-level reporting
• Internal proprietary applications you will integrate with rather than replace

Nice to Haves

• Delivery inside a risk function, or into the risk industry: operational risk, cyber or safety
• Taking early-stage AI prototypes through to high-availability production
• Prior consultancy or client-facing delivery background
• A strong academic record in a numerate discipline
• Postgraduate study in a computational or quantitative field

Projects

The agents you build will make risk decisions a person used to make, pulling from tens of data sources and writing live updates into the register the business runs on. Do it well and you have taken an entire risk function from no engineering capability to AI-first, which is a rare line to have on a CV.

This is the first hire in a programme where the client expects to expand, and the consultancy behind it has more demand for this skillset across its accounts than it has people to meet it. Deliver here and the next thing is bigger.

This is quite a secretive client with valuable IP, so you will learn a lot of the detail only once you are through the door. And the two days a week on site in central London are fixed, not flexible.

AI Engineer | Python, Agentic AI | Finance House – Risk Team