Work¶
Case studies here follow one structure, so they can be compared: the situation, what I personally owned, the constraints, the decisions that mattered and what each cost, what was delivered, what was measured, and what I would now do differently.
Two rules apply to every entry.
Maturity is stated, never implied. A prototype is called a prototype. Where something was not adopted at scale, that is on the page, not omitted from it.
Enterprise work that cannot be shown is not replaced by vague claims. Client identities, product internals and named stakeholders are withheld. Architecture, method and measured outcomes are published, and each page states exactly what is withheld and why. Where the artefact itself cannot be published, the pattern derived from it is — in Methods.
Case studies¶
Prototype · 2026 · CPG
A governed knowledge platform for revenue growth management¶
Institutional knowledge existed across thousands of pages of wikis, BI definitions and spreadsheets — and was not answerable. Ingestion to grounded, cited answer in roughly three months: contract validation with quarantine, four intent-routed retrieval paths, and a competency-question evaluation harness that produced a measured quality baseline before anyone asked to scale it.
Also the harder deliverable: a written account to client leadership of exactly what stood between the prototype and production, and an investment case built on evidence gates rather than a fixed scope.
Further engagements¶
Summarised from the CV. These have not yet been written up as full case studies; the entries below claim only what the CV supports.
| Engagement | Period | What I owned | Outcome |
|---|---|---|---|
| Data platform migration tooling | Mar – Sep 2025 | Led an 8-person team building migration tooling for 70+ downstream engineering teams | 500TB+ across 200+ applications moved off Cloudera to an AWS lakehouse; an AI-assisted code converter reaching 60% automation coverage; reconciliation tooling cutting manual verification by half |
| News editorial platform migration | — | Delivery leadership across a 10-squad, ~70-person programme | Delivered; client commended release smoothness against expectations |
| Retail data warehouse modernisation | — | Delivery leadership, 25-person distributed team | Modernisation delivered; team velocity doubled over the engagement |
| Financial crime risk extraction | — | Built an NLP solution from scratch — BERT and CoreNLP on AWS SageMaker | Risk-signal extraction delivered as a working ML solution |
A fuller record, including roles, dates and certifications, is on the CV.
What is coming¶
An internal pre-sales review platform built for my employer — an eight-dimension analysis pipeline over deal and delivery material — is next in line for a case study. It needs its own sanitisation pass first: it is an internal system, and the same disclosure rules apply to an employer as to a client.