London, UK
Michael
Curran
AI Solutions Consultant
Field service engineer moving into AI consultancy. I build production systems, deploy enterprise infrastructure, and ship AI products independently. Currently at Zenzero, building GrafterAI on the side.
Open to opportunities
Featured project
GrafterAI
A live cross-platform app that automates UK tax tracking for tradespeople. Snap a receipt, Claude and Gemini extract the data. No manual entry. Real users, shipping updates.
React, Capacitor, Supabase, Vercel, Gemini, Claude API
At a glance
Field service engineer. Enterprise infrastructure, Copilot deployments, pre-sales scoping. Sprint Starter Award winner.
Founded and run a football team. More in the Life tab.
University of Liverpool
BSc CSComputer Science with Software Development. Final year project: Digital Wellbeing App with TensorFlow.
Augustus Dons
I started a football team called Augustus Dons. Got the group together, sorted kits, organised matches, and keep it running.
Add the story here. When it started, how many players, your position, best memories.
Liverpool
Originally from Liverpool. University of Liverpool for four years studying computer science. The city shaped most of what I care about.
London
Based in London now. Still settling in.
Where I have been
A proper world map this time — drag to pan, scroll or pinch to zoom, and click a highlighted country for the story.
Countries in colour are ones I have actually been to — I will light up more as I travel.
Right now
Learning
Claude Certified Architect, AB-730, AB-900.
Building
GrafterAI updates, this website, an enterprise RAG pipeline on Azure.
Upcoming
What I am building next
Weekend projects designed to deepen my understanding of AI architecture and build real tools I can talk about in client conversations. Each one solves a genuine problem.
AI Site Report Generator
Weekend 1
A field engineer finishes a site visit and speaks into their phone. The app transcribes, structures, and generates a formatted PDF report with photos, fault descriptions, and recommended next steps. Built for how engineers actually work, not how software people think they work.
Whisper API, Claude API, PDF generation, React Native
LEARNS: Audio transcription pipelines, structured output from unstructured speech, mobile-first AI
MCP Server for ConnectWise
Weekend 2
A Model Context Protocol server that connects Claude to ConnectWise PSA. Ask Claude natural language questions about open tickets, billable hours, client history, and SLA status. The kind of integration MSPs would pay for.
MCP Protocol, ConnectWise API, TypeScript, Claude Desktop
LEARNS: MCP server architecture, API integration patterns, building tools Claude can use natively
Contract Comparison Engine
Weekend 3
Upload two versions of a contract or proposal. The tool extracts key terms, highlights what changed between versions, flags risks, and generates a plain-English summary of what matters. Enterprise legal teams would use this daily.
Claude API, PDF parsing, diff algorithms, document chunking
LEARNS: Document processing at scale, long-context strategies, building trust-critical AI tools
AI Cost Estimator for Enterprises
Weekend 4
A calculator that takes a business use case description, estimates token volumes, compares pricing across Claude, GPT, and Gemini, and generates a cost projection with ROI analysis. The tool I wish I had for every pre-sales conversation.
Multi-model API pricing, React, data visualisation, prompt analysis
LEARNS: Model economics, token estimation, building tools for the sales process itself
Prompt Engineering Workbench
Weekend 5
Test the same input against different prompts, models, and temperatures side by side. Save prompt libraries, track which versions perform best, export results. A tool for people who write prompts professionally, not casually.
Claude API, GPT API, Gemini API, comparison UI, prompt versioning
LEARNS: Multi-model evaluation, prompt optimisation patterns, building internal tooling
RAG Pipeline Visualiser
Weekend 6
Upload documents, watch them get chunked, embedded, and indexed in real time. Then query them and see exactly which chunks were retrieved, their relevance scores, and how the final answer was assembled. Makes the invisible visible.
Azure AI Search, vector embeddings, D3.js visualisation, Claude API
LEARNS: RAG internals end-to-end, embedding models, retrieval debugging, client-facing explainability
What I am looking for
AI solutions consultant or technical pre-sales roles. Somewhere that needs someone who can talk to a client in the morning and ship code in the afternoon.