Senior Principal Architect - AI ML and Cloud Security - AI Architecture

Overview

Senior individual contributor designing the AI solution architectures that GSK's AI Engineering team builds, with design authority over the agentic framework, LLM orchestration, RAG and retrieval boundary patterns, embedding security and API security for AI services.

What you'll do

• Design AI solution architectures covering LLM orchestration patterns, agent design, RAG and retrieval boundary patterns, embedding security, and API security for AI services.
• Validate the agentic framework design before handoff to Engineering and confirm it is buildable, testable and operable.
• Author pattern library entries for LLM orchestration, retrieval boundaries and tool-calling scopes that engineering teams build from.
• Lead architecture reviews for new AI tool proposals and assess them against Cyber Architecture and Solutions standards.
• Contribute to the data fabric architecture and the controls applied at each classification boundary.
• Produce specifications detailed enough that Engineering builds without reinterpretation.
• Apply the binding AI security standards and threat models to every design, escalating where a design cannot meet them rather than diluting them.
• Support use-case feasibility assessment for business and R&D proposals.
• Sit in the agentic framework build stand-ups and answer design questions the same day.
• Review the agent and RAG designs produced by other architects and sign them off or send them back.
• Lead your team in using approved AI tools to improve productivity and ways of working, role-modelling responsible use and following GSK's Responsible AI, privacy and security requirements.
• Own the AI architecture target state and the roadmap towards it for the domain, keeping it current as the agentic framework and the data fabric evolve.
• Apply GxP and data integrity requirements to AI designs so validated systems and the evidence they depend on are not disturbed.

What you'll need

• 15+ years in security or solution architecture with a recognised architecture certification, or 8-12 years in security or solution architecture (domain focus: AI Architecture, agentic framework, LLM orchestration and RAG design, as a senior individual contributor at enterprise scale).
• Agentic framework architecture: agent design patterns, LLM orchestration, RAG and retrieval boundaries, trust boundaries, tool-calling scopes and human-in-the-loop controls, validated as buildable, testable and operable before handoff.
• AI threat modelling and secure AI data flow design: vector store access control, classification, and AI-specific threats (indirect prompt injection, training-data poisoning, model extraction, insecure output handling, excessive agency; OWASP LLM Top 10, MITRE ATLAS).
• Assessing new AI tool proposals against binding AI security standards: model provenance, data residency, classification and third-party risk.
• Producing architecture specifications detailed enough that engineering builds without rework or reinterpretation, with design authority only and build and assurance owned elsewhere.
• Bachelor's degree in a relevant discipline, or equivalent professional experience.

Nice to have

• AI architecture reference pattern library and architecture governance, with Engineering and the vendor building against it.
• EU AI Act alignment and AI regulatory requirements in a regulated industry.
• Azure OpenAI and Vertex AI private endpoint and customer-managed-key patterns, and network designs that keep an inference call inside the EU.
• Experience of designs handed to a third party to operate.
• Cloud AI services architecture: Azure OpenAI, Vertex AI and private model hosting patterns.
• Recognised architecture certification such as SABSA or TOGAF.

Details

• Location: Bengaluru, Luxor North Tower.
• Seventeen roles across the UK, India and the US.
• Typical tools and technologies include LangChain, LangGraph, LlamaIndex, pgvector, Pinecone, Weaviate, LiteLLM, MITRE ATLAS and OWASP Top 10 for LLM Applications.