I architect AI SaaS and automation systems for funded founders and operators with stalled builds.

I've spent the last decade building production systems — full-stack web platforms, SaaS products, multi-tenant infrastructure, and increasingly, AI-native systems. The work spans six verticals: recruitment and HR tech, manufacturing and ERP, voice AI and conversational systems, marketplace and e-commerce, professional services automation, and enterprise workload planning. Thirty-plus shipped products across four continents, most of them still running in production today.

Recently the centre of gravity has shifted to two areas. The first is AI-native SaaS and voice AI: Verdapt — a multi-tenant voice agent platform built on Twilio Programmable Voice, ElevenLabs, and OpenAI Agents with full real-time STT → LLM → TTS streaming, credit wallets, Stripe metered billing, and calendar sync. The hard parts of voice AI are not the prompts; they're the streaming, the multi-tenant isolation, the latency budget, and the billing. I've shipped all of it in production.

The second area is AI automation and multi-system orchestration. GetBoardWise is the proof point — a 5-app monorepo that runs an entire board-placement outreach business: AI-generated personalized messaging (Claude + GPT-4o in parallel, sector classification, decision-pressure extraction), Phantombuster + AdsPower + Oxylabs execution layer, Typeform → HubSpot → AWS Lambda pre-sales pipeline, Slack + Zoom alerting. Lead-to-meeting cycle time dropped 78% (3 days to 16 hours), 99.9% uptime across 12,000+ runs, +19 NPS.

A meaningful slice of my engagements is rescuing software projects that stalled with a previous shop or freelancer. The pattern repeats: a previous agency burned through a budget and didn't ship; a no-code build broke under real load; a founder built 70% themselves and hit a technical wall. Callan Hawkins (UK/Dubai executive search) and Vulcan Jewellery (UK manufacturing ERP) are both rescue-pattern engagements shipped to production on modern stacks in 6–10 weeks. The first time you've been burned, you can't afford a second failed attempt — I take that seriously.

On the deeper-integration end, I work with senior technical founders who've hit a wall on a complex piece — the algorithm, the orchestration layer, the integration depth. WhatNext is an enterprise planning system on a proprietary Urgency-Weighted Fair-Share Scheduling algorithm — real-time Google + Outlook sync, three-level project hierarchy, audit and compliance layers. Those engagements skip the hourly hook and go straight to milestone-based delivery.

I take a maximum of two active builds at a time and one or two advisory retainers. The buyer who hires me is usually one of three: a funded seed-to-Series-A founder who needs senior architecture and execution; an operator at an established SMB whose previous build didn't stick; or a CTO at a growth-stage SaaS team adding an AI product line without an LLM-experienced hire on staff. I work on a modern stack — Next.js, TypeScript, Supabase, PostgreSQL, OpenAI, Anthropic, Twilio, Vercel, AWS — and I move fast because Claude Code, deep prior-art reuse, and a senior delivery team make that possible.