Services
We build applications and systems that use AI where it does real work, and we stay to run them. One accountable team from the first scoping conversation to day-to-day operation.
AI-powered applications
Products with AI at the core, for the desktop and the web. We design the product, write the application and back-end code, choose where each model runs, and prepare the software for distribution.
Typical deliverables
- A native or web application, from first design to a distributable build
- On-device models where privacy requires it, cloud models where the user opts in
- Production pipelines that use generative models, with independent quality checks on the output
AI systems and automation
Agents and pipelines that do operational work inside the tools a business already uses. They work through real interfaces when no API exists, follow written procedures, and stop for human approval at the steps that matter.
Typical deliverables
- AI operators that work a web application through its user interface
- Queued, triggered runs with approval steps and a record of every run
- Multi-step pipelines that combine models with fixed rules and checks
Business systems, built and operated
Systems where correctness matters more than novelty: partner and affiliate programmes, attribution, commission ledgers, QR infrastructure and staff panels. We build them with our AI-assisted engineering method and operate them after launch.
Typical deliverables
- Partner registers, referral codes and links, and attribution rules
- Ledgers that are append-only, with monthly close and refund handling
- Staff panels with roles and two-factor sign-in
- Integrations with the shop or platform you already run
How an engagement runs
-
1. Scope
We agree what the system must do, what it must never do, and how we will know it works. You get a written scope and a quote. We need access to the people who know the process, and to the tools the system will connect to.
-
2. Build
We design and build in increments you can see and test. You get working software, its tests, and a record of decisions. We need timely answers at decision points and test data or a test environment.
-
3. Operate
We deploy, monitor and maintain the system, and change it as the business changes. You get a running system and one contact accountable for it. We need a named owner on your side for approvals.
Every engagement is scoped and priced per project.
How we use AI
We use AI in two distinct places, and we say which one applies to each piece of work.
- In the product. The software uses a model to do its job: recognising speech, generating illustrations, operating an application.
- In the build. We engineer with a team of AI agents: a lead agent and specialists write the code, and separate read-only reviewers check it. Their findings are fixed before release. Decisions about scope, money and release stay with people.
Where privacy needs it, models run locally. Dictorio's speech recognition and correction run on the user's Mac.
Technology
- Cloudflare Workers, D1 and Pages
- Postgres
- Swift (macOS)
- Python and FastAPI
- React and TypeScript
- Local models via whisper.cpp and llama.cpp
- LLM APIs
- n8n workflow automation
- WooCommerce integration
Who we work with
We work best with founders and operating teams of small and mid-sized companies who need a specific system built and then kept running, and who can name the person who decides.
Have a system in mind?
Tell us what you want to build or automate, and where it has to fit. We reply within two working days.