Loading…
Loading…
Selected work
Real-life solutions and leading technology.
From charities to e-commerce to creative AI tooling, these projects show how I approach digital work: practical, inventive, capable, and built for the people who need it.
Over 20 years bringing technology into the real world — offshore, in plants, inside global organisations — turning complexity into solutions that work. I have led teams, designed systems, built software, and solved problems that do not come with instructions. Science, engineering logic, creative vision, and digital innovation, wired with a future-facing mindset.
I did not come up through technology on a straight line — I carved a path through complex, high-stakes industries. For over two decades I have been the person organisations call when problems do not ship with instructions, when new technology must integrate with old reality, or when teams need calm leadership in harsh environments.
My work has taken me offshore onto rigs and production facilities, deep into industrial plants, and into global engineering programmes. I have led multidisciplinary teams, built new service lines from scratch, and delivered digital tools that help organisations understand assets, reduce emissions, optimise performance, and make smarter decisions.
I have always worked where data, software, and physical reality meet — programming, GIS, analytics, and automation to turn messy complexity into clarity. Whether building monitoring systems, digital workflows, or applications that automate what used to take entire teams, the focus is the same: make the complex simple, the invisible visible, and the future usable today.
Alongside that, I have built a parallel practice in full-stack development, automation, and digital product design — Python, JavaScript, React, GIS, and agentic workflows for tools that templates cannot touch.
None of this is theoretical — it is lived, hands-on experience from places where technology has to work, not just look good on a slide.
Today I bring that history to clients who want something original — digital experiences that are visually bold, technically sharp, and unapologetically individual.
Systems, software, and tools that turn industrial-scale complexity into clarity.
Offshore platforms, industrial sites, global organisations — where tech must work.
Full-stack, automation, GIS, and data science in one thread.
Multidisciplinary teams in high-pressure environments.
Modernisation and new-energy thinking long before it was fashionable.
A few representative builds — practical problems met with inventive, responsible technology. Each one started with a real constraint and ended with a measurable result.
How an AI-driven enrichment workflow amplified an already well-built shop and unlocked its next stage of growth.
The client had spent years building a solid ecommerce platform with a loyal customer base and a strong product range. What they needed next was not a rebuild — it was a way to scale product information without scaling headcount.
I implemented an agentic automated workflow that enriched every listing with clearer, richer, more complete information. Sales doubled within eight weeks.
The workflow collects and standardises titles, specs, attributes, and images from the ecommerce backend, then cleans and validates them for consistency, gap-fill, and enrichment-ready structure.
From there it generates enriched content — descriptions, features, benefits, technical detail, SEO metadata, and optional comparisons — on-brand, with quality checks that flag missing data, contradictions, or hallucinations before publish, and automated updates where appropriate without bottlenecks.
We had built a good shop, but we did not have time to enrich every listing. Amelia's workflow gave us the depth and consistency we were missing, and the results were immediate.
Customers buy with confidence when they understand what they are looking at. The workflow delivered that clarity at scale without loading the client team with busywork — it amplified their work. The product catalogue did not change; clarity did.
A three-phase AI editing system that boosts an author's productivity without replacing their creative voice.
A fiction author wanted editing support that worked with their process — not AI rewriting voice, overriding style, or replacing their editor. They needed efficiency and full creative control.
The product vision was a subscription-style author workspace: manuscripts organised as projects, chapters, and scenes; AI feedback stored as comments tied to each section and phase; background queue processing for long works; and version snapshots when a phase completes so the next phase unlocks on a stable baseline.
I designed and specified an agentic three-phase pipeline — developmental editing, line editing, and copyediting — with transparent prompts, structured output formats, a traceable comments log, and audit-friendly change history.
Phases follow professional editing practice. Each phase uses a specialised agent, and the writer accepts, rejects, or revises every point — the goal is never to replace an editor, but to smooth the path from rough draft to intentional manuscript.
Focus on plot, character, pacing, and structure. Feedback targets pacing and flow, character development and motivation, plot logic, structural options where re-ordering helps, and narrative tone and voice consistency — delivered as structured, actionable items (issue type, location, explanation, suggestion) with no wholesale voice replacement.
Focus on style, clarity, and sentence-level improvement. Feedback targets unclear sentences, stronger word choice, redundancy, rhythm and structure, showing vs telling, and dialogue. Suggestions are explicit — original text, issue, suggested direction — so the author stays in control.
Focus on grammar, spelling, and consistency: punctuation, typos, capitalisation and hyphenation, unintentional POV or tense shifts, and alignment to a chosen style guide where configured. Output is correction-oriented and precise — a clean handoff toward a human editor.
This tool is part of my routine. It helps me get from rough draft to polished draft faster, without losing my voice. It has not replaced my editor — it makes me a better writer before the manuscript reaches them.
It respected craft, mirrored real editing phases and feedback shapes, supported process instead of dictating it, and put ethics at the centre — not as an afterthought. The result: higher output without losing ownership, and clearer manuscripts before the editor sees them.
A custom-built, real-time raffle system that powered the charity's largest fundraising event to date.
The charity wanted modern livestream fundraisers with smoother, fairer raffle draws. Manual steps slowed the show and capped growth. They needed something purpose-built, trustworthy, and easy for volunteers — not a generic corporate stack.
Tickets are generated and stored as purchases arrive, with secure randomisation and clear on-camera feedback for fair, transparent draws. Presenter-friendly one-click controls keep the pacing smooth.
Bold visuals read on camera, and the system stays reliable throughout — no lag, crashes, or awkward pauses — with a repeatable setup for future events.
The raffle app transformed our livestream. It made everything smoother, more exciting, and far easier to run. Supporters loved it — our biggest fundraiser ever.
Built around their real workflow, it respected their community personality and removed friction without removing human connection — technology that feels helpful, not corporate.
A multi-modal tool that reads P&IDs from vector files and legacy scans, builds equipment inventories, maps full process flow, and automates mark-ups with agentic precision.
Process teams rely on P&IDs, but drawings live in a chaotic mix of CAD vectors, PDFs, and old scans. Extracting equipment lists, mapping flow, and applying mark-ups is slow and error-prone.
I built an agentic P&ID intelligence system that reads vector and raster drawings, identifies equipment, lines, and instrumentation, builds structured inventories, maps process flow end to end, generates drawing registers, and supports natural-language mark-up requests.
A six-stage pipeline takes drawings from ingestion through to agent-driven mark-ups, with automatic pipeline selection for each format and traceable output at every step.
Vector PDFs, CAD exports, scans, and mixed archives — with automatic pipeline selection.
Tags, instruments, line numbers, valves, control loops, and process boundaries — including legacy symbol conventions.
Equipment lists, line lists, instrument indexes, and tag-to-service relationships — exportable and traceable.
Flow direction, upstream/downstream relationships, boundaries, and interdependencies.
Revision history, titles, system references, and metadata — less spreadsheet archaeology.
Natural-language prompts — highlight pumps, flag missing line numbers, annotate revisions — applied consistently and transparently.
This system transformed how we work with P&IDs. What used to take days now takes minutes, with better accuracy than manual extraction. It is core to our workflow.
It respected the discipline, handled messy real-world drawings, and used agents to amplify human expertise — clarity and speed without compromising integrity. Senior engineers spend their time on engineering, not clerical work.