Product Engineer (AI Team)
- Hiring from
- United Kingdom
- Work type
- Hybrid
- Posted
627,052 remote jobs, straight from company career pages
100% free · New jobs every minute
Show job descriptionHide job description
About Us
Lupa is building a category defining product the industry has never seen before.
We are the AI-native operating system for veterinary practices and pet parents, replacing the fragmented, clunky systems vets have tolerated for years with a single intelligent platform for scheduling, client communication, clinical documentation, and AI-driven care guidance. Practices run more efficiently, vets get back to doing what they love, and pet parents feel more connected to their animals health than ever.
The traction speaks for itself: one of Europe's top 100 AI startups with a team of over 80 people, 10x growth in twelve months, the UK market leader, and now charging hard into the US and Europe. We have $25M in funding, 3M+ pets on the platform, and a buzzing HQ in Paddington, London.
AI is not a side project here. Jerry, our AI teammate for vets, launches this month with clinical decision support, a listening mode for live consults, and hands-free voice next. You will ship weekly to real clinics and see the impact directly.
In this role, you will:
Build Jerry's clinical guideline library: RAG over vet-approved guidelines using hybrid vector and keyword search in Postgres, so Jerry answers clinical questions only from approved sources and cites the exact lines it used
Build Jerry's listening mode: a fast classifier that watches live consult transcripts and triggers Jerry to surface missed charges, tests, or risks in real time
Build Jerry's voice mode: a hands-free, voice-first experience for ambulatory vets on the road, where everything is spoken and any change waits for the vet's approval
Build long-running and scheduled Jerry tasks: turns that keep running after the vet closes the tab, and scheduled jobs that prepare drafts for the vet to approve
Own what you build in production across the TypeScript server and Python AI service: deploys, monitoring, fixes, weekly releases, and a share of on-call
Be a trusted reviewer on AI pull requests and write the design docs for new AI features
Your background looks something like:
4+ years of professional software engineering, including at least a year shipping LLM features to real users in production
TypeScript in production (Node.js backend and React) and Python in production (FastAPI or similar)
Has built with LLM APIs in production: tool calling and agent loops, structured output, prompt and model iteration
Deploys and owns code in production: CI/CD, Docker, a cloud platform (AWS ideally), monitoring and fixing live issues
SQL and Postgres, with git fluency across branches, rebasing, pull requests, and code review
RAG experience: embeddings, hybrid vector and keyword search (pgvector), reranking, citations (nice to have · the first project)
Voice AI experience: real-time voice agents, speech-to-text, telephony integration (nice to have)
Works well with AI coding agents (Claude Code, Codex) and reviews their output critically
As a person, you:
Are a product engineer first · you care about what ships and how it performs in front of real users, not just whether it works in a test environment
Own your code end to end, from design doc to production, and do not consider a feature done until it is monitored and stable
Can explain the code you wrote, including the code your AI tools wrote for you
Are comfortable in a small team with a big scope, working directly alongside vets, product, and the wider engineering org
Are excited to work from our London WeWork office most of the week
Value working with people who are kind, ambitious, and pragmatic
What does success look like in 6 months?
Shipping to production every week and contributing to on-call
Jerry's guideline library v1 is live for pilot clinics: it answers only from approved guidelines, cites exact lines, and passes the clinical team's test questions
Listening mode is piloted with at least one clinic, with every suggestion tracked to show the revenue it recovers
Long-running and scheduled tasks are in production and a voice-first Jerry has been tested with an ambulatory clinic
