Senior AI Engineer

Our client is a rapidly scaling UK deep-tech company delivering advanced AI capability into some of the most consequence-heavy environments: defence and national security, critical national infrastructure, and enterprise clients whose decisions carry real weight. They are hiring two Senior Applied AI Engineers to take ownership of the agentic AI systems underpinning their platform, on a contract basis, with genuine influence over how those systems are architected and delivered.

This role suits an engineer who designs and ships systems rather than assembling pre-built components - someone who cares more about how a system performs once it’s live and under pressure than how impressive it looks in a demo.

What You’ll Be Doing

• Architecting and building multi-agent AI systems end-to-end, combining frameworks such as LangGraph or Haystack where they add value with bespoke orchestration, state management and tool-calling logic where a framework won’t cut it.
• Writing production-grade code with reliability, latency and graceful failure handling designed in from the start rather than bolted on afterwards.
• Embedding LLMs and vision-language models into live reasoning, search, summarisation and task-execution workflows, backed by solid retrieval, memory and guardrail design.
• Deploying across cloud, on-premises and fully offline/air-gapped settings, adapting your approach to each environment's security and connectivity constraints.
• Taking end-to-end ownership of production pipelines, from raw data ingestion through to live inference.
• Building the monitoring, evaluation and guardrail tooling that shows how a system is actually performing in the field.
• Setting technical standards for agentic AI within the team, establishing the patterns other engineers follow and leaving clear documentation behind.

What We’re Looking For

• Robust software engineering, including developing LLM-powered or agentic systems, and a track record operating at senior/lead level.
• Strong engineering fundamentals - code that’s clean, tested and built to last, not just prototypes that work once.
• Commercial track record building and owning multi-agent or agentic AI systems in live production settings, as an engineer rather than an integrator.
• Strong Python skills, comfortable working with LangGraph, LangChain, Haystack or similar, and equally comfortable dropping below the framework when that’s the right call.
• Real-world experience deploying AI/ML systems into production, beyond notebooks and proof-of-concepts.
• Confidence across inference performance, data pipeline design, state/memory handling, and tool or function calling.
• Experience building retrieval and search over very large, multi-modal datasets (multi-terabyte scale, spanning text, imagery, telemetry and sensor feeds), including indexing and embedding at scale.
• Comfortable with Docker, Git and cloud infrastructure, ideally AWS.
• Understanding of secure deployment approaches: air-gapped, on-premises or sovereign cloud.

Nice to Have

• Multimodal reasoning and/or edge or offline AI deployment experience.
• Kubernetes experience (EKS or OpenShift).
• MLOps background: evaluation, monitoring, reproducibility.
• Observability for agentic systems, covering drift, agent behaviour and performance.
• Familiarity with agent-to-agent orchestration protocols or the Model Context Protocol (MCP).
• Awareness of secure-by-design frameworks (ISO 27001, NIST, OWASP).
• Background in defence, national security or other regulated sectors.
• Open-source AI/ML contributions.

Why Apply

• Real technical ownership: shape the architecture, not just work through a backlog.
• Work at the leading edge of agentic and generative AI, on systems people genuinely depend on.
• Join a fast-scaling business with a strong chance of the contract extending or evolving further.
• Flexible, hybrid working arrangement.
• Strong day rate reflecting seniority and scope.

This is a contract engagement operating outside IR35, offered on an initial 6-month term with a strong likelihood of extension.