Support Copilot
Claude-powered agent grounded on your help docs and order data. Drafts replies inside Intercom / Zendesk with confidence scores, escalates the rest. Cuts tier-1 handle time without losing the human voice.
Freelance automation engineer building AI agents, workflow automations, data pipelines and document extraction for founders, ops teams and operators who refuse to keep paying a human to copy/paste at 2am.
Free audit on the call · No deck, no fluff
Custom GPT/Claude agents that read your docs, answer customers, qualify leads, write reports — wired into your existing stack with proper guardrails and logging.
n8n / Make / Zapier or custom Node + Python jobs. Connect your CRM, sheets, email, billing, slack, dashboards — kill the 'someone manually does this every Monday' tasks.
ETL, scrapers, API integrations and dashboards. Pull the data you care about, clean it, store it, visualise it — without a 6-figure data team.
Email-in pipelines that read invoices, contracts and PDFs with vision LLMs, validate against your schema, and push clean rows to Xero, QuickBooks, your warehouse — wherever they need to be.
Claude-powered agent grounded on your help docs and order data. Drafts replies inside Intercom / Zendesk with confidence scores, escalates the rest. Cuts tier-1 handle time without losing the human voice.
Email-in → vision LLM extraction → schema validation → Xero / QuickBooks sync. Anomalies flagged for human review with one-click approval. Replaces a full-time AP clerk's worst week.
Scrape → enrich → LLM-qualify → push to HubSpot with a fit score and a personalised first-line. Sales team wakes up to a warm pipeline; ops doesn't touch a spreadsheet.
Working under NDA on a couple of client builds. Happy to walk through architecture and decisions on a call.
20-min call. We map the workflow, find the wasted hours, agree the smallest version of the system that makes a dent.
Flat-fee scope. One page. Inputs, outputs, success metric, what it explicitly does NOT do. No change-order theatre.
I ship a working version in days, not months. You see progress on a staging URL with logs. We adjust fast.
Deployed, monitored, documented. You get a Loom walkthrough, the repo, and 14 days of warranty bug fixes.
A single workflow or agent, scoped tightly, shipped in 5–7 days. Best for a clearly defined bottleneck.
End-to-end build of a serious workflow: AI + data + integrations + dashboards. Two weeks of focused work.
Fractional automation engineer. Weekly slot for new builds, fixes and improvements. Cancel any time.
Fixed-fee pricing. No hourly games. If we can't agree a scope that fits your budget, we won't start.
Sprint projects ship in 5–7 days. A full Build is two weeks. You see a working staging URL within the first 72 hours — not weeks of silence.
No — most clients don't. I deliver a documented system with a Loom walkthrough so anyone non-technical can operate it. The Operator package exists specifically for teams who don't want to own the maintenance.
Almost certainly. I prefer modern tools (TypeScript, Python, Postgres, n8n, Claude) but I work inside whatever you already have — Zapier, HubSpot, Salesforce, Notion, Airtable, Sheets, Slack, etc.
Every build ships with logging and alerts. You get 14 days of warranty fixes free, and an Operator retainer if you want ongoing peace of mind.
I build with guardrails: retrieval grounding, output validation, prompt-injection defences, scoped credentials and audit logs. No shortcuts on safety.
Sometimes, for the right kind of project. Talk to me on the call.
UK based. Work with clients globally. Currently accepting 2 new projects for Q3 — first come, first scoped.