Senior Advanced Research Engineer in Penarth

We are at the forefront of a new era in enterprise AI — one defined not by model capability alone, but by the infrastructure, memory systems, and routing intelligence required to make autonomous AI agents trustworthy and commercially viable at scale. Our Data & AI practice brings together more than 45,000 professionals helping clients design, deploy, and govern AI systems across regulated industries. Our applied research function sits at the intersection of frontier AI research and production engineering — investigating the foundational challenges that will determine whether enterprise agentic AI succeeds or stalls.

You sit at the boundary between AI systems research and production platform engineering. You investigate hard, open problems in agentic AI — and you close the loop: turning research findings into engineered prototypes, then into platform-ready capabilities that real workloads depend on. You are a strong Python engineer who can move fluently between an experiment and a well-structured service or SDK module. You write research artefacts and production code in the same week, and you understand why both matter.

The Work:

• Applied Research & Innovation: Investigate active innovation frontiers in agentic AI systems — for example, agent memory and knowledge persistence architectures, model selection and inference routing strategies, autonomy and goal-anchoring control planes, and long-horizon task reliability. Design and execute rigorous benchmarking and evaluation methodologies scoped to production-relevant agentic task profiles — covering dimensions such as tool use, structured output generation, multi-step reasoning, instruction following, and failure recovery. Investigate efficiency and scalability frontiers — such as inference cost reduction, context management at scale, and retrieval architecture design — that determine whether agent workloads can be served commercially on attainable hardware. Contribute to external publications, technical reports, and conference submissions that establish thought leadership and build the evidence base for client and platform decisions.
• Translational Engineering: Translate research findings into production-grade implementations: engineered Python services, Node.js/TypeScript SDK modules, or platform-integrated components that other engineers and agent workloads depend on. Build well-defined provider interfaces and pluggable backends for research components — memory stores, retrieval layers, routing modules — so that experimental implementations can be iterated on and swapped independently of the platform code that depends on them. Prototype and validate platform-level capabilities — such as inference routing policies, memory management layers, or agent control mechanisms — and carry them through from experiment to integrated, observable system component. Instrument research prototypes with observability from the start — distributed tracing, cost accounting, and latency metrics — so findings are reproducible and platform integration is low-friction.
• Platform Contribution & Integration: Work alongside platform engineers to integrate validated research capabilities into production systems — contributing well-tested, documented Python and Node.js/TypeScript code through standard engineering workflows including code review, CI, and schema validation. Identify platform gaps surfaced by research experiments — missing APIs, insufficient observability, constrained interfaces — and raise them as concrete, scoped engineering proposals. Ensure that research-derived capabilities meet production standards: correct error handling, sensible defaults, documented contracts, and test coverage appropriate to their risk profile.
• Collaboration & Communication: Work closely with platform engineers, product managers, and enterprise architects to align research priorities with real client deployment blockers and platform roadmap needs. Communicate research findings, architectural trade-offs, and prototype results clearly to both technical peers and non-technical stakeholders — in written artefacts, design reviews, and client-facing sessions. Mentor junior engineers and researchers on experimental methodology, translational engineering practices, and production-quality code standards. Travel may be required for this role. The amount of travel will vary from 0 to 100% depending on business need and client requirements.

Here’s what you need:

• Bachelor’s degree (or equivalent minimum 12 years work experience, or minimum 6 years' work experience with Associate’s degree) in Computer Science, Computer Engineering, or a related field.
• 5 years of experience with Python and/or Node.js/TypeScript, building and shipping production backend services, research prototypes, or AI/ML systems.
• 5 years of hands-on experience with AI or ML systems — such as large language models, agent frameworks, inference serving, or retrieval and memory architectures.
• Bonus points if you have:
• 6+ years of engineering experience across both research and production contexts, with a demonstrated ability to ship research into running systems.
• Deep experience in at least one area of applied AI systems research — such as agent memory and knowledge management, inference efficiency and model routing, agentic evaluation methodology, or long-horizon task and autonomy research.
• 3+ years of applied research with a track record of translating findings into platform-integrated or published artefacts — prototypes, open-source contributions, internal frameworks, or peer-reviewed papers.
• Hands-on experience with async Python (e.g. FastAPI, asyncio), containerisation and Kubernetes, vector and relational databases, and distributed tracing instrumentation (e.g. OpenTelemetry).
• Familiarity with modern AI agent framework ecosystems and agent communication protocols — the specific tools matter less than the ability to work across multiple frameworks and evaluate them critically.
• Master’s or PhD in Computer Science, Computer Engineering, or a related field is strongly preferred.