Corvus
A persistent personal AI agent — semantic memory on local Ollama embeddings, hybrid Python + LLM execution, and cron-driven daily briefings. Privacy-first, runs on a Mac Mini.
Four ventures you can engage, and the systems behind them. I build what I recommend — these are real, running, and most of them I use every day.
Making AI actually work inside organizations.
VisitAI & Claude Code courses for senior career-changers.
VisitPractical resilience for people exiting the default.
VisitYour career, as an interactive AI.
VisitA persistent personal AI agent — semantic memory on local Ollama embeddings, hybrid Python + LLM execution, and cron-driven daily briefings. Privacy-first, runs on a Mac Mini.

A cross-AI, JSONL-based memory layer that gives Claude, ChatGPT, and local models shared, persistent context across tools.

A privacy-first, self-hosted LLM stack — Ollama, OpenWebUI, and an MCP bridge, auto-starting and always on.
An operations dashboard for a fleet of AI agents, with ADHD-aware UX — a "Three Things" focus widget, brain-dump triage, and a signal-filtered activity feed.
A multi-agent org structure — a chief-of-staff agent plus ops, sales, and product agents coordinating across every venture.
A fully self-hosted media stack — Jellyfin, Radarr/Sonarr automation, VPN kill-switch, mergerfs storage, Tailscale remote access — documented so anyone can run it.
A privacy-first home voice system — Home Assistant + Wyoming with on-prem speech, routing deterministic commands locally and complex queries to a local LLM.
A DIY-designed residential solar + 32 kWh battery system — full energy modeling, structural and permitting work, phased rollout, at a fraction of installer cost.

A neighbor-to-neighbor tool-sharing platform — AI-assisted cataloging, 170 users across 20 communities, now pivoting to library partnerships for distribution.
That's exactly what Velocity Partners does.
Start a conversation