Senior Full Stack Engineer

Who you are

• AI raises the floor. We raise the ceiling - that's our engineering vision. The floor is everything that happens without a person doing it by hand: implementing, verifying, reviewing, deploying, correcting, rolling back, around the clock. The ceiling is what's possible, and raising it is our responsibility
• Accountable - utterly accountable, and driven by results. You'll be joining a small team who are keenly motivated to improve our systems and our output every day
• You'd be the overarching orchestrator - you set direction, wield the tooling exceptionally well, and own the result. You'll shape what the harness looks like and drive change classes up the ladder, while shipping product to housing associations who are using this every day
• Verification discipline - meaningful tests, contracts, data invariants, evals, observability. You understand that when code becomes cheap to produce, the definition of "correct" becomes the durable asset - and you'd enjoy being one of the people who builds ours
• Senior full-stack, production-hardened - years of shipping complete systems across frontend, backend and database, and operating them once they're live. You've been on the hook when it broke
• Raising others - you review well, mentor without taking over, and leave the codebase easier for the next person and the next agent
• Architectural judgement - you make structural decisions autonomously and can defend them: service boundaries, data modelling, where complexity belongs and where it absolutely doesn't
• Prioritisation under pressure - you can tell a fire from a smoke alarm and say no to the wrong work
• Deep TypeScript, React and Next.js - advanced patterns, rendering strategies, state management; you have opinions and reasons for them
• Client-facing - comfortable on calls with housing associations, translating their workflows into system design and technical decisions back into plain English
• Databases you'd trust with money - schema design, query optimisation, migrations, integrity constraints, history. You think in invariants, not just tables
• Fintech, proptech or regulated-domain experience - genuinely useful here, though the domain is learnable and the judgement isn't
• Cloud and infrastructure - AWS in production, comfortable with infrastructure as code (we use Pulumi), environments, secrets and blast radius
• You direct AI at a high level and catch what it gets subtly wrong - this is a core assessed skill, not a bonus. We want to hear where you've overridden a model and why you were right to. You don't need to have built agent infrastructure before; you need the engineering judgement to build it and the appetite to learn fast
• You want to invent how we build, not just what we build - the AI infrastructure this company runs on is still being designed, and there's no playbook to copy. We want someone who'd rather design it than inherit it
• Domain correctness is non-negotiable to you - in fintech, plausible-looking maths that's wrong is worse than no feature. You test the money paths like someone will audit them, because someone will
• Product-first - you argue about what to build and for whom before you argue about how. We hire engineers who are product people first, deliberately

What the job involves

• The market - real estate is the largest asset class in the world, and a home is the most important asset most people will ever own. Stairpay is transforming the infrastructure that runs affordable housing in the UK and beyond, bringing automation and efficiency to a sector that has had very little of either, and making sure residents feel the benefit first
• The platform - it manages the purchasing journey, uses AI for document ingestion and regulatory checks, gives residents an app through which they manage their home, and provides the financial layer for large institutional landlords collecting millions in rent every month. Shared ownership is where we started and where we're strongest: the hardest tenure, and the one the incumbents handle worst
• One record underneath all of it - one property, one lease, one ledger, one resident identity, so nothing is ever reconciled between systems, because nothing is ever in two places
• The domain - statutory clocks, lease terms that differ per property, money maths where being slightly out is a regulatory problem. That's the moat
• The autonomy ladder - defined classes of change earn their way up: agent implements, human reviews the code → agent implements and proves it, human approves the deploy → ships itself, monitored, with automatic rollback. Promotion is earned per class, on a track record. Money movement, authentication and destructive data migrations never pass the middle rung: a human approves those forever, and in our sector that cap is something we sell rather than something we admit
• The engineering underneath it - a full local stack that spins up in minutes, isolated parallel environments, E2E tests everywhere (they're cheap to write now, and written well they're fast to run), QA agents running their own testing passes alongside them, infrastructure entirely in code, and signals from Sentry and PostHog that tell the platform what to fix next
• Where the humans sit - every promotion frees human attention for a layer higher: what's worth building, architecture, taste and UX judgement, customer understanding, and designing the evals and guardrails that define what "correct" means. The frontier keeps moving, and that's the point rather than a caveat
• Own a domain end-to-end - concept → architecture → implementation → deployment → operation → iteration. You make the architectural calls; nobody is going to hand you a spec
• Orchestrate, verify, and take responsibility - direct agents across implementation, review and verification, then bring the fundamentals to bear on what comes back. The human is the director; AI is the tool. You are accountable for the output either way
• Earn autonomy for change classes - design the tests, gates and evidence that let a class of change move up the ladder, then make the case for the promotion with the track record to back it
• Fix the class, not the instance - every production bug becomes a test, and where possible a lint rule, type or guardrail that makes the whole category impossible
• Build the foundations that make it all possible - full local stack in minutes, isolated parallel environments, E2E tests littered across the product, QA agents running their own passes, everything in code, proven recovery. In our sector every one of these is also a sales asset
• Work directly with clients - join calls with housing associations, validate solutions with real users, demo and iterate
• Set the technical direction with us - you'll have real influence on the roadmap, the stack and how the team works. We're small enough that your opinion changes what happens

The application process

• AI is assumed, not banned - our process is built for how we actually work
• Apply - CV or LinkedIn / WTTJ profile, plus a brief note on why Stairpay
• Initial screening call (~30 mins) - two-way fit and career trajectory
• Async technical test (~2 hours, your own time) - review a pull request on a real repo of ours, identify the three issues you'd block the merge on, fix them with tests, and record a short Loom walking through your reasoning, including where you used AI and where you overrode it. The repo comes pre-set-up, so setup friction doesn't eat into your two hours
• In person: problem-solving session and bar-raiser - a live, multi-layer technical case, then a conversation about your trajectory, the achievements you're proudest of and how you work
• In person: final interview and meet the team - values, motivation and mutual fit with our CEO and Head of Product, plus time with the wider team
• References and offer

Benefits

• 25 days holiday plus bank holidays
• Work from anywhere for 2 weeks a year (July, August, December)
• Gym in the office
• Generous stock options
• Monthly team socials
• Latest equipment, and the AI tooling budget to go with it
• Work remote 1-2 days a week