Services
Nine things we build properly
Most projects are a mix of two or three of these. Every one of them ships for a one flat monthly fee, deployed on your accounts, with the code yours from the first commit.
01
MVP development
You've had the idea for six months and a slide deck instead of a product.
Not a prototype, not a clickable mock. Software real customers can sign up for and pay for, small on purpose and complete where it counts.
- Auth, data model, billing and admin as needed
- Deployed on your accounts with monitoring
- A demo every Friday from week one
- Two weeks of post-launch support
- Next.js
- Postgres
- Stripe billing
- Auth
02
AI agents & LLM products
"Add AI to it" isn't a scope — and a chatbot that makes things up is worse than no chatbot.
Most AI projects die because nobody wrote down what the model is actually responsible for. We start there: one job, one measurable output, and a defined answer for what happens when the model is wrong.
- A grounded assistant or pipeline running on your data
- Evaluation set so you can prove it works before launch
- Guardrails, fallbacks, and logging for every call
- Cost-per-request measured and tuned
- OpenAI / Anthropic APIs
- pgvector
- RAG
- Structured extraction
03
Full-stack development
Your team runs the business out of spreadsheets and Slack DMs because no one's built the internal tool yet.
The internal tool your team is currently faking with spreadsheets, or the customer portal that has been on the roadmap for a year. Next.js and TypeScript, deployed on your infrastructure.
- The app, responsive and accessible
- Roles and permissions that match how you work
- APIs and integrations to your existing systems
- Error tracking and analytics from day one
- Next.js
- TypeScript
- REST / GraphQL APIs
- Admin panels
04
API development
Your API works until someone outside your team tries to use it without asking you a question first.
Whether the customer is your own frontend, a partner integration, or a paying developer, the interface they touch is the product. We design it, version it, document it, and keep it stable.
- REST or GraphQL API, versioned from day one
- Authentication, rate limiting and quotas
- Generated docs and a client SDK
- Contract tests so a change can't silently break callers
- REST / GraphQL
- OpenAPI
- API keys & quotas
- SDK generation
05
Infrastructure setup
Deploys are a ritual only one person can perform, and nobody finds out it broke until a customer says so.
Most early products run on somebody's laptop config and a deploy nobody wants to touch. We set up the accounts, pipelines and monitoring properly once, on your infrastructure, so shipping stops being an event.
- Cloud accounts and environments set up in your name
- CI/CD pipeline with preview deploys
- Monitoring, logging and error alerting
- Backups, secrets handling and a written runbook
- Vercel / AWS
- Docker
- CI/CD
- Monitoring & alerts
06
Product design
The design looked great in Figma and fell apart the moment real data went into it.
Not a pretty mockup that falls apart the moment it meets real data. Interface design done against the actual states — empty, loading, error, too much text — by the people who then build it.
- Screen designs for the real states, not just the happy path
- A small component system your app can grow into
- Responsive behaviour decided, not left to chance
- Handover straight into the build, because it is the same team
- Figma
- Design systems
- Tailwind
- Accessible components
07
Ecommerce
Your store works fine until a customer does something the platform's defaults never expected.
Platform defaults get you ninety percent of a store. We build the last ten — the custom checkout rule, the catalogue sync, the integration that connects the shop to everything else you run.
- Storefront or custom checkout work
- Catalogue and inventory sync
- Integrations to fulfilment, ERP or CRM
- Live launch with rollback plan
- Shopify
- Stripe
- Custom checkout
- Catalogue sync
08
Web scraping & data pipelines
The data you need lives on someone else's website, and copy-pasting it every Monday doesn't scale.
Public pages turned into clean structured rows, delivered where your team already works. Built to survive the markup changes that break every scraper eventually.
- Extractors for the sources you name
- Scheduled runs with retries and change alerts
- Delivery into Sheets, Notion, a warehouse, or a webhook
- A written note on what is and isn't lawful to collect
- Playwright
- Scheduled jobs
- Postgres
- Sheets / Notion sync
09
Workflow automation
Someone on your team does the same manual process every week and calls it "just how it's done."
Every company has a manual routine nobody questions because it has always been done by hand. We take one, automate it end to end, and make it tell you when it breaks.
- The process automated end to end
- Failure alerts into Slack or email
- An admin view so a non-engineer can re-run it
- Handover doc for whoever owns it next
- Node.js
- Webhooks
- Slack / email alerts
- Internal tools
Next step
Not sure which one you need?
That’s what the call is for. Bring the problem and your deadline — we will tell you which of these it actually is, or that it isn’t one at all.