Skip to main content
New: Try our AI Chatbot assistant — click the chat icon in the bottom-right corner to get started.
Our work

Hard problems.
Production systems.
Lasting results.

Every engagement below shipped into a live environment and left behind a system the organisation could maintain without us.

Filter
Insurance
Security & Governance

Row-Level Security that runs itself — across every client.

A major insurer was configuring Power BI Row-Level Security by hand for each client. Fragile, inconsistent, impossible to audit, and entirely dependent on whoever touched it last.

The problem

Security rules were built individually for each client — copy-pasted, manually adjusted, and never documented. Adding a new client meant a developer's time. An audit meant a week of archaeology. A mistake meant a breach.

Our approach

We built an automation layer on Microsoft Fabric backed by a centralised SQL registry that holds all client security definitions. A single PySpark pipeline reads the registry and applies RLS across every client, every time, in one execution. New clients are added by updating the registry. No code changes required.

What changed

The team configuring security by hand stopped doing it the day we deployed. New clients onboard overnight. Security reviews now pass without a remediation list.

Microsoft FabricPySparkDelta LakePower BI REST APIAzure Key VaultEntra ID

“Security configurations that took days now run overnight. The team hasn't touched a deployment script since.”

Head of Data · Lancashire Insurance
Multi-sector
CI/CD & Automation

Power BI deployments that run like software releases.

Manual Power BI deployments across four environments. Reports breaking in production. No version history. Entirely dependent on whoever last touched it.

The problem

Every release was a manual process — a developer moving files between workspaces, running scripts by hand, hoping nothing broke. There were no gates, no rollback, and no audit trail. When something went wrong in production, the only option was to fix it there.

Our approach

We built a full Azure DevOps CI/CD pipeline with declarative JSON workspace configurations and custom PowerShell modules. Every environment — development, test, UAT, production — is defined in version control. Deployments are triggered automatically, workspace membership is managed programmatically, and approval gates prevent anything reaching Finance or Actuarial without sign-off.

What changed

Releases that used to require a developer babysitting a process now run unattended. Errors stopped reaching production. The team stopped receiving late-night calls about broken reports.

Azure DevOps (YAML)PowerShellPower BI REST APIAzure Key VaultService PrincipalsGraph API
AI Venture
AI & Agentic Development
Live

SoiKio — institutional investment research, fully automated.

An investment research platform built from scratch and shipped with Claude Code. Every architectural decision, every deployment config, every integration designed and delivered through prompting.

The problem

Institutional investment research means weighing macro conditions, fundamental analysis, valuation, credit risk, technical signals, ESG factors, and regulatory exposure simultaneously, for every asset under consideration. No single analyst can do this at scale. Most platforms either automate one dimension or charge enterprise prices for the full picture.

Our approach

We designed and shipped a full stack platform that coordinates 16 focused AI analysts in parallel, each covering a distinct research dimension. A fan-out workflow via n8n Cloud distributes the research task, agents run concurrently against live market data from Polygon.io, Yahoo Finance, and Alpha Vantage, and results are combined into investment memos. Built entirely with Claude Code: architecture, infrastructure as code, data pipelines, and front end.

What changed

A production platform generating institutional quality research in hours rather than weeks. One experienced architect, working through AI, shipping production software.

Next.js 14TypeScriptSupabasen8n CloudCloudflare R2Polygon.ioClaude CodeVercel
Visit SoiKio →
E-Commerce Venture
Built with AI
Live

Curated Lagos — a premium Nigerian gifts marketplace, built with AI.

A production e-commerce marketplace for premium Nigerian goods, designed and shipped with Claude Code.

The problem

The Nigerian gifting market has no credible, curated online presence for premium products. Diaspora customers wanting to send quality gifts have no trustworthy platform, just inconsistent marketplaces with no editorial voice and unreliable fulfilment.

Our approach

Designed and built a full stack marketplace from the ground up using Claude Code. Product catalogue, checkout, fulfilment workflows, and brand identity. Every decision from layout to deployment pipeline made by an experienced architect who knew what production software requires.

What changed

A live marketplace serving customers, built without a traditional agency, without a large team, and in a fraction of the time. One architect with the right tools, shipping a complete product.

ReactCloudflare PagesFirebaseClaude CodeGitHub CI/CD
Visit Curated Lagos →
Work with us

Your problem belongs
on this page next.

Tell us what's not working. We'll be straightforward about whether we can fix it.