Lead Technical Architect (AI) in South Shields

Job Description \n About Scrumconnect Consulting \n Scrumconnect Consulting is a multi-award-winning digital consultancy, recognised for delivering impactful and innovative technology solutions across UK government departments. n We are passionate about user-centred design, agile delivery, and building digital services that make a real difference; and we're now scaling that expertise into large, high-stakes AI adoption programmes across the public sector. \n Overview \n We're looking for a Lead Technical Architect (AI) to act as the technical authority for a full-scale AI Operating Model and Responsible AI Governance Framework at a central government department. This is a large, high-visibility AI adoption and scale-up programme covering systems and services handling high-volume, sensitive public sector data. \n You'll work as part of a collaborative "Rainbow Team" alongside civil servants, with a mission that goes beyond delivery: build sustainable internal capability so the department can eventually operate and govern its own AI capability without long-term reliance on external suppliers. \n Key Responsibilities \n Architectural Leadership & Integration Design and implement enterprise-grade AI platforms - including semantic search, Retrieval-Augmented Generation (RAG), and broader generative AI/agentic capabilities - integrating cleanly with the department's complex Legacy and multi-cloud environments. \n Responsible AI Governance Embed end-to-end responsible AI controls across the full life cycle: alignment with the Algorithmic Transparency Recording Standard (ATRS), NCSC "Secure by Design" principles, and active mitigation of algorithmic bias. n Risk & Value Management Navigate trade-offs between value, risk, pace, and quality using HM Treasury Orange Book and Technology-Organisation-Environment (ToE) frameworks. Identify and resolve systemic risks via Joint Risk Registers. \n Zero-Dependency Handover & Capability Uplift Co-deliver within blended agile teams. Drive knowledge transfer using the OKUA (Ownership, Knowledge, Understanding, Awareness) framework and "Docs-as-Code" practices so internal staff can independently operate and govern AI capability long after the engagement ends. \n Environmental Sustainability (Green AI) Design low-modality, energy-efficient AI architectures aligned with "Circularity Performance Requirements" (Buy Better, Use Better, Use Longer), supporting Net Zero 2030/2040 commitments. \n Required Skills & Experience \n Mapped to the Government Digital and Data (DDaT) framework: \n \n \n Technical Design Throughout the Life Cycle (Expert) : technical designs for high-risk, high-impact, high-complexity systems; n \n Making Architectural Decisions (Practitioner): medium-to-high-risk design decisions spanning multiple domains; n \n Architect for the Whole Context (Practitioner): tracking emerging AI and technology trends; n \n Strategy Design (Practitioner): defining architectural principles, AI patterns, and strategic vision aligned to wider government objectives; building implementation roadmaps. \n \n Community Collaboration (Practitioner): proactive networking, resolving team-dynamics issues, using Agile health checks to strengthen the multidisciplinary delivery team. \n \n Alongside the DDaT competencies, given the scale of this programme, hands-on depth across the modern AI engineering stack is essential, not just RAG in isolation: \n \n \n Semantic search, vector/embedding infrastructure, and retrieval architecture design \n \n LLM orchestration and agentic frameworks (eg LangChain/LlamaIndex -style tooling, multi-agent patterns) \n \n LLM evaluation, guardrails, and red-teaming tooling (hallucination detection, safety/bias testing) \n \n LLMOps/MLOps: model versioning, deployment pipelines, monitoring, and rollback for AI services in production \n \n Fine-tuning and parameter-efficient adaptation techniques where appropriate to use case \n \n Data engineering for AI: ingestion, cleaning, and lineage across Legacy and multi-cloud sources \n \n Infrastructure-as-Code and CI/CD practices adapted for AI workloads, with strong observability \n \n Secure-by-design engineering appropriate to sensitive public-sector data (identity, access control, data protection) \n \n Desirable Experience \n \n \n AW