Aull-Stack Developer (AI-first, spec-driven)
About Us
Our SaaS platform runs scheduling, crews, payments and sales for cleaning companies across the US, and is becoming a multi-vertical field service platform. A legacy system (Node.js, MSSQL) serves today’s customers. Its successor, a multi-tenant platform on AWS (Node.js, PostgreSQL, Angular, Kubernetes), already runs the Sales module and takes on the core of the business next: jobs, scheduling, crews and payments, rebuilt from the ground up in 2027.
How we build
Implementation is cheap now; correct design is not. Every developer here works with AI coding agents (Claude Code) as the main way code gets written. Your value is upstream and downstream of the agent:
— Product rules arrive as signed specifications with numbered rules and acceptance tests, and a clickable prototype of the end-state product. You turn them into technical design: data model, state machines, API contracts, the order of increments.
— You write the ADRs and the work breakdown agents build from, and you review everything that comes back as if a junior wrote it: line by line, against the spec.
— You prove behaviour before anyone sees a screen. Business lifecycles are graph documents executed by our entity state engine, and a simulator plays synthetic companies’ working days against the platform on a virtual clock, so a rule is verified in simulation first.
— You own what ships: release, production monitoring, and the fix when reality disagrees with the spec.
What you will work on
**Now:** finish the Sales release due at the end of 2026 (multi-service quotes, per-branch sales tax, duplicate merge and archive, payment checks at booking, automation timing across US time zones) while the remaining customer companies move onto it. You also take over live services: the Sales/CRM backend and UI, Typesense search, the data sync between the legacy and new databases, and the AI assistant’s data tools.
**Next:** the new Jobs application, built in layers, foundations first:
— Company and branch settings, users, and role-based access across branches
— Services and subscriptions that generate real jobs across a planning horizon
— Scheduling by day routes: day, week, month and map views, drivers, trainees, qualifications
— The provider app, an installable web app for crews on their phones
— Payments and invoicing, payroll and earnings, the customer portal, communications
— Shared platform pieces they all consume: the entity state engine, file storage, reporting
What we look for
— You design before you prompt. Given a business rule, you can name the entities, states, invariants and failure cases, and say what must be atomic and what can converge later.
— You read code more critically than you write it, and you catch the plausible but wrong change an agent makes with confidence.
— You write well: specs, ADRs, tickets with acceptance criteria, review comments. Clear, short, unambiguous English.
— You think in contracts: APIs, events, idempotency, retries, migrations of live data.
— You test by evidence: automated tests, simulations, production checks. Not by “it works on my machine”.
— You are fluent with AI coding agents in daily work, and you know where they fail.
— You can work on both sides: backend services and the Angular UI on top of them.
— You care about the user’s day. Our users are office managers and cleaning crews, not engineers.
Stack
TypeScript end to end. Backend: Node.js 24, Express, PostgreSQL with TypeORM, RabbitMQ, Typesense, WorkOS. Frontend: Angular in an Nx monorepo, Angular Material. Testing: Vitest, Cypress. Platform: AWS EKS, ArgoCD, Sentry, Grafana. Legacy: Node.js and MSSQL stored procedures, which you will read as the source of behaviour you migrate. Integrations: Stripe, Twilio, SendGrid, Google Maps, MCP for the AI assistant.
Nice to have
— State machines, workflow engines or event-sourced systems in production
— Scheduling, dispatch, payroll or billing domains
—
