I'm Will — I build AI automation and websites for small businesses, and I run the whole stack on my own hardware: an 8× RTX 3090 inference rig serving Qwen, GLM, and Flux. No cloud bills, no rate limits, no excuses.
- 🔭 ResearchOS — an AI analyst that researches products and writes you a decision, on local GPU
- 📊 AutoInvoice — voice-or-text → invoice business OS for multiple companies
- 🤖 eve.center — an autonomous agent that sells domains end-to-end, no human in the loop
- 🦾 Building a tendon-driven robot arm from scratch — actuators, controls, the works
Everything ships. Everything runs on hardware I own.
Languages
AI & Automation
Web & Infra
- 🔭 ResearchOS — Ask "best printer under $2k for production parts"; it runs a real research pipeline and writes a scored decision. FastAPI + Vane + local Qwen.
- 📊 AutoInvoice — Describe a job by voice or text, get an invoice. Multi-company transaction tracking with an MCP tool layer.
- 🤖 eve.center — Autonomous AI sales agent: chats, qualifies, finds a domain, takes payment via Stripe, deploys a site. Zero humans. (source)
- 🚀 Nova-Rig — My 8× RTX 3090 local inference rig: vLLM + ik_llama.cpp serving 200B-class models, plus self-hosted search, scraping, and voice. The backbone every other project runs on — zero per-token cost.
Inference: 8× RTX 3090 · Jetson Orin 64GB · vLLM + llama.cpp
Models: Qwen 3.5 122B · GLM 4.7 · Flux · Chatterbox TTS
Stack: Next.js · FastAPI · Postgres · Docker · Caddy · systemd
AI: Claude · local agents · LangGraph
