AI Platform Engineer

About Homeprotect

We founded Homeprotect on a simple principle – to provide protection to people underserved by the home insurance industry.

We’re experts at insuring people who want to build or buy a unique property, renovate, rent a home or leave it unoccupied. We can cover people who travel a lot, run a business from home or even collect rare treasures. We protect people who love living in a listed building, having a view of a river or sunbathing on a flat roof.

We can do this because our smart tech enables our customers to get an instant, online quote to cover a huge range of complex needs and our UK customer support teams are on hand to provide information and support when only a real person will do.

Our insurance has already empowered hundreds of thousands of people to protect their homes and the things they love. But we’re not content with being the leading provider of what’s called ‘non-standard’ insurance. We believe there’s no such thing as ‘standard’. We’re all unique and we all deserve home insurance designed with our individual needs in mind – something our combination of real people and smart tech enables us to do.

Our simple promise? Whoever you are and wherever you live, with Homeprotect we’ve got you covered.

Our customers come from all walks of life and so do our people. We’re a small but perfectly formed team, made up of insurance industry specialists alongside a diverse mix of technologists, data scientists and customer and marketing champions from all sorts of industries and backgrounds. We all bring our individual expertise, an appetite for innovation and a shared ambition to empower people to protect their homes and the things that they love.

Working in a fast-paced environment where change happens regularly is how we do things. But we also recognise that you’ll do your best work when you have the right balance, which is why we’ve fully embraced hybrid working, giving our teams the flexibility to choose the right location and working hours for them. We focus on the outputs of work, not where it’s completed, with an expectation that colleagues spend around one day a week in the office (or a minimum of four days per month) to connect and collaborate.

There will be times when getting together in one location makes sense, but day to day, our teams have the freedom to decide where to work, and we trust them to balance business needs with their own preferences.

Most companies think their culture is great, however at Homeprotect, we have the proof to back this up. We have been recognised externally as a Great Place to Work for the last six consecutive years

Sound good? Read on to find out more about joining our team…

• Design, build and operate AI‑first products that deliver measurable value, shaping roadmaps with the Head of AI and AI Product Manager.
• Develop and deploy production‑grade AI agents for end‑to‑end claims handling on GCP using Google’s ADK; integrate securely with policy/claims/CRM systems via REST/JSON APIs.
• Create and run agent evaluation and monitoring: offline/online tests, accuracy/completeness/safety metrics, A/B experiments, cost/latency dashboards.
• Integrate AI agent functionality with applications via MCP servers. Build and configure tools to ensure reliable application integration.
• Build robust RAG pipelines and ingestion: text normalisation, chunking, metadata/ACL propagation; vector search in GCP.
• Embed security and compliance (GDPR, PII redaction, secrets, RBAC/IAM, audit trails); document/demonstrate solutions with Claims Ops; support go‑live and BAU.

We would love to hear from people with the following skills and experience:

• Strong Python for backend services/SDKs; REST/JSON APIs, auth (OAuth2/service accounts), testing (pytest), async and resilience patterns.
• Hands‑on with LLMs/embeddings (OpenAI, Gemini, Llama, Claude), prompt and tool/function calling, RAG design; vector databases.
• GCP expertise:
  Vertex AI, Model Garden, Big Query, Cloud Run/Cloud Functions; practical experience with Google’s ADK.
• Document/data handling:
  Apache Tika, text normalisation, chunking/dedup; metadata/ACL propagation; storage/caching with…