Work

Systems in production, written down

Client names are withheld because most of this work is under agreement. Everything described here was built and run in production, and every number was measured. More detail under NDA on request.

Prior engagements

Where the patterns were proven

Nearly four decades of engineering, the last ten-plus years focused on AWS. The service lines grew out of these engagements.

Case 03 · 2022 – 2025Developer-experience platform · 500+ engineers

Cloud provisioning as a product, for five hundred engineers

Several years with a fintech’s developer-experience team as it went public. Product teams declared cloud resources in a simple YAML schema, an internal developer platform provisioned them in CI/CD, and a React portal showed each system’s pipelines, status, metrics and logs.

Architected the abstractions teams actually used, for S3, ElastiCache and OpenSearch. The OpenSearch work moved help-desk search onto a highly available multi-index platform and made it much faster. Also designed how specialist teams (machine learning, networking, security) provision their own resources inside the core platform.

Engineers served500+
ScopeGo, AWS CDK v2, Kubernetes
LineDeveloper experience & platform
Case 04 · 2019 – 2021National streaming platform · launch scale

A single-instance monolith made ready for a national launch

Consulting principal on a major streaming service: catalog, live and on-demand scheduling, partner metadata export, real-time ad breaks and worldwide manifest delivery. Then the same platform again for an Olympic summer games.

The central webapp ran on one hand-built instance. We split it for horizontal scale, moved functionality into services, rebuilt the infrastructure in Terraform with reviewed, automated applies, and built CI/CD from scratch. We also ran load testing ahead of a worldwide sporting event and treated cloud spend as an engineering target.

Cloud spend−72% in 11 months
ArchitectureMonolith → horizontal + services
LineRelease reliability & decomposition
Case 05 · 2016 – 2019Ed-tech marketplace · later public

Cloud foundation from zero, then fifty services and a marketing platform

Three roles in three years, up to director. The legacy systems ran on a third-party layer over AWS: no control of resources or spend, no access to critical logs, and someone else’s support desk between the company and its own outages.

We built a secure cloud foundation from scratch (twelve accounts, private networks, shared provisioning and CI/CD tooling), and penetration testers found no infrastructure vectors. All five legacy systems moved in without customer impact. The monolith then became more than fifty services and a hundred serverless functions.

In the final year, as principal architect, built a paid-marketing platform that enforced ad and keyword integrity across about 4.2 billion keywords, against an ad budget of roughly a fifth of company spend.

Decomposition50+ services · 100+ functions
Pen test0 infrastructure vectors
Ad platform~4.2B keywords
Case 06 · 2014 – 2016National retailer · digital office

An ETL orchestration platform over terabyte on-premise data

Built and ran an orchestration platform that moved terabyte-scale on-premise data into cloud Postgres and served it through a horizontally scaled REST API with Redis caching. Pipeline tasks loaded as plugins, so any stage could be extended without touching the orchestrator.

PatternPluggable pipeline tasks
ScopeNode · Python · Postgres · Redis
LineArchitecture & data platform
Case 07 · 2008 – 2014Global software company · enterprise platforms

Enterprise platforms at a scale where milliseconds are policy

Six years at Microsoft. As engineering lead on the display advertising platform, re-engineered the UX tier, two middle-tier database abstraction services, and API contracts over five backend stores holding billions of campaigns: load time for the largest accounts went from over ten minutes to under two seconds.

Before that, as global design lead in the OEM organisation, took worldwide license management from paper to digital and built a declarative UI rules engine that let business partners change logic without a release. It won the company’s IT Pro Innovate award.

Performance10+ min → <2s
RecognitionIT Pro Innovate award
LineArchitecture & performance

This is the level of detail you get on your own systems: written down, measured, handed over.

Start a project