Job description
is building out its internal platform capabilities, and AI is a growing part of that. The role combines architectural ownership with hands-on delivery: you will design, build, deploy and operate the platform, including its infrastructure and DevOps. Reporting to the Director of Product Security, you will set up this function as its first member and make the early technical decisions that the future team will build on.
As Platform Architect, you will own the architecture of the platform that more than 1,000 employees rely on, including their access to AI tools such as Anthropic Claude, OpenAI and future providers.
### Responsibilities:
- Build the AI platform architecture: the reference design for how LLM capabilities are accessed, governed, monitored, and extended across the organisation.
- Own every architectural decision for the AI platform, and document, justify, and stand behind each one.
- Evaluate and select new models, tools, and frameworks as the market moves.
- Deliver the platform as a service, so teams can move their agents into governed cloud infrastructure.
- Govern Anthropic (Claude API, Claude Code, Anthropic Enterprise), OpenAI, and future providers, covering API access controls, credential management, model version policy, cost allocation, and acceptable use standards.
- Own technical relationships with Anthropic and OpenAI at the enterprise level.
- Build cloud infrastructure for AI workloads on AWS and GCP: Bedrock, SageMaker, Vertex AI, Cloud Functions, and supporting services (S3, Cloud Storage, IAM, and monitoring on both platforms).
- Deploy and securely configure n8n, and build automation workflows connecting LLMs to Jira, Confluence, Slack, Google Workspace, HR systems, and proprietary data sources.
- Design and secure MCP servers and gateways connecting AI agents to internal services.
- Build Python integrations for AI-powered internal tooling: RAG pipelines, agent frameworks, prompt libraries, and internal AI assistants.
- Deliver hands-on: you write the code, build the workflows, and deploy the tooling. This isn't an advisory role.
- Build data privacy and security architecture for LLM usage in a regulated environment, including prompt classification, output controls, and audit trails aligned with FCA, CySEC, ASIC, and DORA.
- Build observability across the full stack: token consumption, cost per team, latency, error rates, and policy violations across cloud platforms and direct API providers.
- Build an AI adoption framework of standards, patterns, and onboarding materials that let teams across Technology use AI safely and consistently.
- Run internal enablement through workshops, documentation, and direct support for teams moving from ad-hoc AI use to governed, repeatable processes.
- Report AI platform status, risks, and roadmap to Technology leadership
### Requirements:
- 8+ years in engineering, with at least 2 years building LLM-based systems in production, not prototypes or POCs.
- Comfort operating as the first person in a function: you set your own priorities, unblock yourself, and deliver without a team around you.
- Production-grade Python: tooling that others depend on, not scripts you hand off.
- Deep, hands-on experience with the Anthropic and/or OpenAI APIs: function calling, tool use, streaming, RAG, prompt management, token optimisation, and model versioning.
- Hands-on experience with agentic platforms, MCP servers, and MCP gateways, including designing flows between AI agents and internal services.
- n8n: deploying, securely configuring, and maintaining multi-step automation workflows at organisation scale.
- Solid AWS for AI workloads: Bedrock, SageMaker, Lambda, S3, IAM, and CloudWatch, deployed and operated in production.
- Solid GCP for AI workloads: Vertex AI, Cloud Functions, Cloud Storage, IAM, and Cloud Monitoring.
- Hands-on DevOps capability to deploy and run what you build.
- Architectural thinking: end-to-end system design, clear trade-off documentation, and the ability to defend decisions to senior stakeholders.
- Security engineering for what you build: secrets and credential management, key rotation, and least-privilege IAM across cloud and SaaS.
- Data protection and auditability by design: classifying what data can reach which model and provider, controlling and logging prompts and outputs, and building audit trails that hold up to a regulator.
- Fluency in the LLM threat surface, including prompt injection, data exfiltration through model outputs, insecure tool and agent use, and model supply-chain risk, with working knowledge of a recognised framework such as the OWASP LLM Top 10 or NIST AI RMF.
### Nice to have:
- Fintech, brokerage, or financial services background.
- DORA or ISO 27001:2022 knowledge as it applies to AI system risk.
- Familiarity with the EU AI Act and other AI regulatory requirements, as well as AI-specific risk and governance standards such as NIST AI RMF or ISO/IEC 42001.
- LLM orchestration frameworks: LangChain, LlamaIndex, or similar.
- Hands-on experience with LLM guardrail and DLP tooling: Bedrock Guardrails, Vertex AI safety filters, or prompt-firewall and PII-redaction tools.
- Experience building internal developer platforms or AI centres of excellence.
- AWS Solutions Architect Professional, Google Professional ML Engineer, or equivalent.
