I'm an AI engineer with 6+ years of software experience. I build AI systems that do real work — qualifying leads, checking invoices, answering customers from your own documents — and every one of them is live on the internet right now for you to try.
WHAT I BUILD
• LEAD QUALIFICATION AI — every inbound enquiry scored against evidence and written straight to your CRM. For a resort client, response time went from hours to under 60 seconds.
• INVOICE & DOCUMENT AI — reads invoices before payment, extracts the data, and halts anything suspicious for human approval. Nothing gets paid on an AI guess.
• AI SUPPORT CHATBOT — trained on YOUR knowledge base, answers with inline citations, escalates to a human when confidence is low. Handled 90+ real visitor conversations on a live client site.
• CONTENT & WORKFLOW AUTOMATION — n8n + Claude API pipelines connecting Notion, HubSpot, QuickBooks, Slack, Gmail, Sheets. One pipeline cut publishing time from 30 minutes to 2 minutes per post. If it has an API or a webhook, I can wire it up.
• FULL-STACK WEB APPS — Python/FastAPI back ends, Next.js/React front ends, PostgreSQL/Supabase, deployed with Docker on AWS or Vercel. I own everything from data model to DNS.
WHY MY AI SYSTEMS DON'T EMBARRASS YOU
Most AI demos fall apart on real data. Mine ship with a PUBLIC evaluation suite — pass rates, cost per decision, and documented failure cases against labelled test sets. My systems are built to answer "not found" instead of inventing something plausible, and to hold consequential decisions for a human instead of auto-approving. You get verified numbers, not promises.
PROOF — all live, click and try them:
• verdict.----------------------.com — agentic lead qualification with CRM writes + public evals
• ledgerguard.----------------------.com — invoice AI with human-in-the-loop financial controls
• provenance.----------------------.com — RAG document Q&A with exact cited sources
• upskill-course-online.vercel.app — learning platform with citation-backed RAG assistant
• ----------- — enquiry automation, hours to under 1 minute
• ----------------------.com — full portfolio + case studies
TECH: Claude API, OpenAI API, Python, FastAPI, RAG, pgvector, structured outputs, prompt-injection defenses, n8n, Zapier, Next.js, React, TypeScript, PostgreSQL, Supabase, Redis, Docker, AWS, Terraform, Vercel, GitHub Actions, HubSpot API, QuickBooks API, Slack API, webhooks
AVAILABILITY: Full-time or part-time. Comfortable with US, AU, and UK hours. Stable fiber + backup internet.
TELL ME THE TASK YOUR TEAM HATES DOING.
Send me a message describing it and I'll tell you — free — whether it can be automated, roughly how long it takes, and what it would cost. If it can't, I'll say so.
-----------
Verified
