Hi, I'm Daz.

I've been writing code since 2000 and shipping companies since 2008. Originally from England, I came to the US on an O-1 "Extraordinary Ability" visa and became a citizen. Today I'm in Dallas, Texas, and I work out of FluxAI US Inc., the studio I co-founded.

Across two decades I've built and run e-commerce companies (Harland Corp, Redu Group), been CTO of a $30M hair-and-beauty network (Beauty&Hair), co-founded one of the earliest voice-AI shopping assistants (CODY, 2015 - before "voice AI" was a category), and built a social publishing platform with 0.9B annual impressions (SocialX).

For the last three years I've been heads-down on AI - architecting, coding, and deploying AI-native products end-to-end. Three of them are in production right now: a meeting intelligence platform, a phone-and-web AI receptionist, and a private-cloud autonomous SDR.

Full-stack AI engineering. The whole loop - frontend, backend, AI pipelines, infra. Python and TypeScript across the stack, and the AI plumbing underneath: custom RAG, vector indices, prompt orchestration, structured LLM output, cost-tracked LLM calls, MCP tool use.

Production infrastructure. I run GPU servers I built myself - CUDA, vLLM, model serving, batch and queue systems. Container orchestration on my own hypervisor, mesh-VPN networking, dual-WAN failover, tunnel-based ingress. I've personally served hundreds of thousands of generated images and full-duplex voice traffic from this stack.

Real-time voice. WebRTC, NAT traversal, ICE candidate gathering, custom real-time voice pipeline, STT/LLM/TTS stitched into sub-three-second latency. Phone and browser, end-to-end.

The boring stuff that makes AI production-grade. Webhook idempotency, rate limiting, OAuth (with separate scopes for separate integrations), tenant isolation at the queryset layer, RFC 8058 unsubscribes, dry-run modes, observability. Most "AI projects" hand-wave this; mine don't.