I ship LLM systems into real customer-facing production — not demos. Sales agents, real-time voice AI, autonomous agents, and the self-hosted GPU infra underneath them.
📞 Call the number above and you'll talk to an AI receptionist I built — it screens the opportunity and texts me. The call itself is a demo of my work.
Claude sales closer on a live Shopify store + Facebook Messenger. Live cart awareness, session memory, dynamic product UI, and a hallucination guard that blocks fabricated prices before they reach a customer. Real paying customers.
Real-time voice agent: Deepgram STT → streaming LLM → ElevenLabs TTS, low-latency per-sentence. Caller memory, 6 industry personas, ROI/booking tool calls, missed-call text-back.
llama.cpp serving a 27B model across a Tesla M40 + RTX 3060. Custom CUDA build for Maxwell hardware; tuned to 19.97 tok/s (1.90× over naive split) by benchmarking layer placement.
An autonomous agent that health-checks systems, ranks marketing opportunities, and publishes content on a schedule with no human in the loop. First autonomous post verified in production.
One-shot AI business launcher: intake form → website + 3-variant ads + 7-day social calendar + video commercial + jingle, for ~$1.40 per client.
Food-delivery platform (Laravel + Flutter, 183 restaurants, AI menu photos, voice ordering) and a live client marketing site with AI imagery + Kling hero video.
Every number below is checkable. Call the phone number. Visit the stores. Curl the APIs. I don't make claims I can't prove.
Marion, OH. Built and operate the entire technical stack: live storefront, AI sales agent, voice AI, autonomous marketing, self-hosted PBX, payments, and self-hosted multi-GPU LLM inference. Every project above is deployed here. Real customers, real revenue.
I go into the messy real problem, wire the model into the actual workflow — CRM, storefront, phone system, payments — and put guardrails on it so it's safe in front of a paying customer. That's the Forward Deployed / AI Automation Engineer job.
Yes. I'm based in Marion, Ohio and work remotely with clients nationwide. Every project on this page was built and deployed remotely. I'm comfortable working async with Slack/Discord, or jumping on calls when it matters.
Anything that puts an LLM, voice model, or autonomous agent into a real customer-facing workflow — sales agents, voice receptionists, autonomous operators, internal tools, and the GPU/infra underneath them. I don't build demos; I ship production systems that handle real customers and real money.
That's the core of what I do. I wire AI into the systems you already use — Shopify, Stripe, your CRM, phone system (Twilio/Asterisk), Facebook, email. The "Rick" sales agent and voice receptionist above are examples: they plug into live storefronts and phone lines, not sandboxed playgrounds.
Both. I run self-hosted LLM inference on my own GPU cluster (Tesla M40 + RTX 3060, Proxmox) for clients who need data privacy or want to avoid per-token API costs. For high-capability tasks I use Claude, GPT-4, or OpenRouter — whatever fits the latency/cost/quality tradeoff.
Most projects go from kickoff to production in 1-3 weeks. I built the LuxuriousComputers AI sales agent, voice receptionist, autonomous operator, and full Shopify storefront as a solo engineer. Speed comes from owning the whole stack — no hand-offs, no waiting on other teams.
It depends on scope. A focused integration (adding an AI chat agent to an existing site) is a different engagement than building an entire autonomous system from scratch. Email me at nodnarbrox@gmail.com with what you're trying to do and I'll give you a straight answer — no sales calls, no discovery workshops.
Scheduled agent loops, health checks, growth analysis, and Facebook Reels — all without human intervention.
Real-time STT → LLM → TTS pipeline on Asterisk with sub-second latency and missed-call recovery.
Framework for finding, building, and deploying AI automations — with real HVAC client code and cost comparison.
Tips and guides from our team: