Join Admiral Business as a Principal Engineer - Agentic Platform in a senior individual-contributor role based with flexible hybrid working in Cardiff. You will shape shared AI agent capabilities, connect language and voice experiences to business systems, and drive technical direction across teams.
Requirements
Track record of designing, building, and operating production software across multiple teams or business domains.
Practical experience taking LLM-powered or agentic systems into real use, including tool calling, orchestration, and evaluation.
Strong software engineering skills in Python, TypeScript, or comparable languages.
Experience with distributed systems, APIs, event-driven integration, and cloud infrastructure on AWS.
Understanding of agentic system failures, autonomy, human oversight, security, and reliability.
Skills & Responsibilities
Python and TypeScript: Build robust backend services and agentic logic to connect language models with enterprise business systems.
Agent Orchestration Frameworks: Design reusable orchestration, tool execution, and conversation state capabilities for multi-step workflows.
Distributed Systems and AWS: Engineer reliable execution through durable state, bounded loops, retries, and cloud integrations.
Evaluation and Observability: Establish automated tests, regression checks, and release controls for models, prompts, and tools.
Technical Leadership: Coach engineers, create reference implementations, and guide technical direction across teams.
Potential transition routes: People from Senior Software Engineer, ML Platform Engineer, or Backend Tech Lead roles may bring strong distributed systems, API development, and software engineering skills, alongside meeting experience with LLM or agentic systems requirements.
Job description
How do you build AI agents that can complete useful work, connect to business systems and know when to hand over to a person?
At Admiral Business, we’re building the shared capabilities that make this possible across customer, broker and operational journeys. We’re looking for a Principal Engineer to shape our agentic platform, connecting language and voice experiences to the systems where work happens.
This is a senior individual-contributor role with substantial hands-on engineering. You’ll take systems from early experiments into production, set technical direction across teams and help others build on the platform. Your influence will come through technical expertise, practical delivery and making the engineers around you more effective.
Why this matters
Our ambition is one reusable platform that supports many insurance products, brokers, partners and channels. Agentic capabilities will help us improve how customers access services, how brokers work with us and how our teams complete everyday tasks.
You’ll work alongside platform teams, building on shared services and integrations to give agents the context, tools and controls they need to complete business workflows.
Success means agents that complete permitted tasks reliably, reduce operational effort and improve customer outcomes, with clear human oversight and useful handovers when people need to take over.
The foundations you build should make each new workflow easier to launch, giving teams shared orchestration, knowledge and evaluation capabilities they can use across the business.
What you’ll be doing
Own technical direction for the shared agent capabilities that sit across our insurance platform, connecting engineering priorities to customer experience, operational capacity, cost, risk and speed to market.
Work with Product, Operations, Data and Security to identify valuable workflows, understand how they could fail and agree measurable acceptance criteria, choosing the right balance of conventional workflows, model reasoning and human decisions.
Build reusable capabilities for agent orchestration, model access, tool execution and conversation state, enabling agents to gather information, select tools, execute permitted actions and verify outcomes across multiple steps.
Work with platform teams to expose policy, claims, payments, CRM and data capabilities as secure, reliable tools for agents, including systems such as Root, Five Sigma, Stripe, Salesforce and Snowflake.
Engineer reliable execution through durable state, bounded agent loops, timeouts, retries and reconciliation, so interrupted workflows can resume without duplicating business actions.
Build knowledge and retrieval approaches that respect access permissions, preserve sources and make missing information, uncertainty and conflicting evidence clear.
Develop voice and conversational journeys using platforms such as ElevenLabs, designing for interruptions, response times, transcription errors, identity checks and handover to people with useful context.
Establish evaluations, regression checks and release controls for changes to models, prompts, tools and orchestration. Monitor decisions, actions, task completion, escalation quality, latency and cost, with staged rollout, fallback behaviour and rollback.
Build appropriate access controls and data boundaries, protect against prompt injection and untrusted content, and require approval for consequential actions.
Remain hands-on with critical components and production problems, creating shared libraries, reference implementations, delivery pipelines and documentation. Coach engineers and technical leads, support hiring and help teams build and operate agents more independently.
What you’ll bring
You’ll have the engineering depth to build critical parts of the platform yourself and the judgement to guide how teams use agentic systems in practice.
We’re looking for:
A track record of designing, building and operating production software across multiple teams or business domains, with evidence of meaningful business impact.
Practical experience taking LLM-powered or agentic systems into real use, including tool calling, orchestration, evaluation and production support.
Strong software engineering skills in Python, TypeScript or comparable languages, with the ability to contribute directly to complex implementations.
Experience with distributed systems, APIs, event-driven integration and cloud infrastructure on AWS or an equivalent platform.
An understanding of how agentic systems fail, including unreliable reasoning, incorrect tool use, poor context, partial execution and changes in model behaviour.
Sound judgement about autonomy, human oversight, security, reliability and cost, including when a conventional service or workflow is the better solution.
Experience establishing automated tests, evaluations, observability and release practices for systems whose outputs can vary.
The ability to turn ambiguous problems into practical designs and sequenced delivery, explain trade-offs clearly and align senior stakeholders.
Evidence of improving the effectiveness of multiple teams through coaching, shared capabilities and technical leadership.
Experience with agent orchestration frameworks such as LangChain or LangGraph, voice platforms such as ElevenLabs, retrieval-augmented generation or MCP would be useful, as would experience in regulated industries.
Our wider stack includes Terraform, AWS, React and Snowflake. You don’t need every tool we value sound engineering judgement and the ability to learn quickly.
Why Admiral Business?
Commercial insurance has remained largely unchanged for years. At Admiral Business, we’re doing things differently.
We’re combining the strength and stability of one of the UK’s leading insurers with the pace, ambition and mindset of a business that’s building for the future. By bringing together underwriting expertise, technology, data and customer insight, we’re creating a smarter Commercial proposition that’s designed around better outcomes for brokers, customers and our people.
Joining Admiral Business means becoming part of a collaborative team where expertise is valued, curiosity is encouraged and everyone has the opportunity to make an impact. Whether it’s building stronger broker relationships, using insight to influence commercial decisions or helping develop the next generation of our Commercial proposition, you’ll see the impact of your work every day.
What you'll get in return
We believe exceptional people do their best work when they're trusted, supported and rewarded. Alongside a competitive salary, you'll receive a benefits package designed to help you grow your career, share in our success and maintain a healthy work life balance.
Highlights include:
A generous share scheme, giving you the opportunity to become a shareholder and share in Admiral's success.
33 days' holiday, including bank holidays, increasing with service, plus the option to buy or sell up to five additional days.
Ongoing learning and development, leadership programmes and opportunities to grow your career across Admiral Group.
Flexible hybrid working, with most colleagues spending one or two days a week in our Central Cardiff office.
A people first culture recognised as a Great Place to Work for more than 25 years, including awards for Wellbeing and Women.
To explore our full range of benefits, including wellbeing support, healthcare, family friendly policies and financial benefits, visit our Benefits Page.
Disability Confident Leader
As a Disability Confident Leader, we’ll invite a fair and proportionate number of applicants with a disability or long-term health condition who opt into the Disability Confident scheme and meet the essential requirements of the role to the first stage of our selection process.
If you need any adjustments or support with your application or during the recruitment process, please email Daniel.Cavill@admiralgroup.co.uk.