What I build, and what done looks like.

01 / ai agents and automation

One agent, one job, the right tools, a gate where actions matter.

An agent is a program that reads, decides and acts on your systems within limits you set. I build single-purpose agents that do one job reliably, with retrieval over your documents, guardrails, and an approval gate wherever an action has consequences. Automation sprints chain several of them with shared error handling.

# evals/report.txt (excerpt)
suite: invoice-triage v3        cases: 42
pass: 41  fail: 1  threshold: >= 95%   PASS
failed: case 17 "missing PO number"  -> escalated to human (expected)
cost/run: $0.19   p95 latency: 2.4 s

02 / copilots and rag pilots

A private assistant that answers from your data and shows its sources.

Retrieval-augmented generation done properly: your documents and systems, your permissions, an assistant that cites what it used and says when it does not know. Run as a pilot with real users and a go/no-go memo at the end.

03 / aios, an ai operating layer

One governed place where models, data, tools and workflows meet.

The layer between your people, your systems and the models: which model runs, who may ask it what, which tools it may touch, who approves an action, and what is logged. The Blueprint designs it in two weeks; the Foundation puts the first workflows live on it. How AIOS works, with a demo →

04 / websites

Hand-built, fast, accessible, yours.

No page builder, no plugin stack, no theme licence. Static or headless CMS, dark and light themes, a performance budget promised in writing, and the source handed over with a deploy pipeline. This site is built the same way.

# budget.yaml (what the proposal promises)
lighthouse_mobile: >= 90 on performance, accessibility, best practices, seo
js_total_gzip: <= 30 kb        # no framework, no tag manager
lcp_4g: <= 1.5 s   cls: 0
fonts: self-hosted, subset, preloaded
third_party_on_load: none

05 / mobile apps

iOS and Android from one codebase, in the stores.

One codebase (Flutter or React Native), the backend it needs, and the store submission handled end to end, including the review issues nobody budgets for. Store accounts are opened in your name.

06 / apis, integrations and data platforms

Systems that talk to each other, documented.

A REST or GraphQL API with an OpenAPI spec and a Postman collection, one system finally connected to another, or the data layer under all of it: schema, migrations, pipelines, a warehouse only when it earns its place, and the dashboards on top.

-- schema.sql (excerpt)
create table orders (
  id          uuid primary key default gen_random_uuid(),
  customer_id uuid not null references customers(id),
  status      text not null check (status in ('draft','paid','shipped')),
  created_at  timestamptz not null default now()
);
create index on orders (customer_id, created_at desc);   -- the query the dashboard runs

07 / cloud and devops

Infrastructure as code, deploys you can trust, bills you can read.

Your cloud account, described in code, with a pipeline that deploys on merge and a preview for every change. Logs, metrics and alerts routed to you, an on-call runbook, and a cost review with a written savings list. Included in most builds; sold alone when that is what you need.

# .github/workflows/deploy.yml (excerpt)
on: { pull_request: {}, push: { branches: [main] } }
jobs:
  preview:  { if: github.event_name == 'pull_request', runs-on: ubuntu-latest, steps: [test, build, deploy-preview] }
  release:  { if: github.ref == 'refs/heads/main', needs: [test], steps: [build, migrate, deploy, smoke-test] }
  rollback: one command, documented in the runbook

08 / audits and technical consulting

A straight answer on your stack, your plan or your vendor's proposal.

The One Bit Audit scores AI readiness (data, capability, operations, governance) or a technical system (performance, security, accessibility, architecture, cost) and ends with a prioritised plan. The Second-opinion Session is ninety minutes and a written summary. The retainers put a senior engineer in your leadership loop.

# scorecard.md (excerpt, scale 0–4)
data          | availability 3 · quality 2 · access 1 · lineage 0
capability    | skills 2 · tooling 3 · budget clarity 1
operations    | documented 1 · approval paths 2 · measurement 0
governance    | policy 0 · risk appetite 2 · oversight 1
first build   | the one use case that scores high on value and feasibility

add-ons · flat prices, on any build

Add-ons.

Converted prices are approximate. You are billed in USD.

Source-code handover and a documentation pack are included in every offer at no charge. Regulated-industry work adds 20 to 35 percent for procurement, security review and compliance documentation.

what i will not sell you

Five things that are not on this page on purpose.

• Agent washing: a chatbot called an agent
• A platform you do not need
• A rewrite when a fix will do
• An unscoped retainer
• Hourly billing without a cap