Product engineering
Accounts, billing, permissions, dashboards, notifications, exports. The parts a product needs before its hundredth customer and never gets in a prototype.
Services
Most studios sell a discipline. We sell finished software and the judgement around it — which is a longer list, and the only way to be useful to a business that has already bought the wrong thing once.
A — Build
Accounts, billing, permissions, dashboards, notifications, exports. The parts a product needs before its hundredth customer and never gets in a prototype.
Three fixed shapes — Launch, Catalogue, Commerce — priced in person-days with the CMS included rather than rented back to you monthly.
API design, webhooks with idempotency, two-way sync, SSO, and migrations that finish instead of running forever in the background.
Re-platform the part that is a dead end while the rest keeps running. Parity is proved by driving the old system and the new one through the same actions.
The engineer left, the deploy is manual, nobody knows why the queue backs up. We take it over, write down what is true, then fix it in order of risk.
Tokens, components and the accessibility rules that stop a rebuild in year two. Built to be handed to an in-house team, not held hostage.
B — Applied AI
Ingestion, chunking, embeddings, hybrid search and citation — permission-aware, so people only retrieve what they are allowed to read.
An assistant with the context of the screen it sits on and a bounded set of actions, with the human confirmation where the action is irreversible.
A procedure you already follow, turned into software that runs it: tool calls against real systems, guardrails, idempotency, and an audit trail per step.
The queue of manual decisions, replaced by a scored classifier with a confidence threshold and a fallback to a human where the score is low.
A labelled set from your own history, a score, a regression gate in CI, and drift alarms — the artefact that lets you change a prompt without fear.
Routing, caching, batching, distillation and quantisation, with cost per request known in advance and a ceiling per tenant that cannot be exceeded.
C — Assurance
Review is the highest-leverage thing a senior engineer does, and the thing most likely to be sold as a scan. A scanner tells you a pattern exists. A review tells you whether the pattern is reachable, and proves it with the path that reaches it.
We read at the source: the file, the line, the call path, and the reason it matters. If a finding does not reproduce, it does not go in the report — a shorter list that holds up is worth more than a long one that does not.
D — Data
Metric definitions are signed before anything is built. A dashboard nobody trusts is worse than no dashboard.
Pipelines that surface margin by client, order velocity, inventory turns and anomalies before they become incidents.
Move the data, not the mess. Mapping, deduplication, reconciliation reports and a rollback path that has been tested rather than described.
Stakeholders see sprints, blockers and deliverables without adding a meeting, written up weekly in plain language.
E — Advise
Reference architecture, integration map, risk register and a fixed-price build estimate. Two weeks, and it ends in a number rather than a document.
Which three use cases in your business are actually worth automating, what the data situation really is, and what each would cost to run per month.
Twenty hours a month of architecture ownership, hiring guidance, roadmap execution and board-level technical reporting.
Workshops and code review for your own engineers — what to delegate to a model, what never to, and how to prove the difference in CI.
We deploy, configure and run the agent against your stack. You own no infrastructure and get the audit log. Monthly, with a response window in writing.
Not sure which fits?