> whoami
20 years in the trenches of security. now i run a full self-hosted AI fleet on budget Intel Arc silicon — local models, autonomous agents, a memory brain. all of it mine, none of it in the cloud. the useful pieces ship here.
- Neuralis — a persistent brain that remembers across every tool & agent. hybrid semantic + graph recall, self-correcting.
- Sentinel Prime — autonomous desktop & fleet agents. goals in, work done — guardrails clamped on money & anything outbound.
- AURORA-X — one glass pane over the fleet: voice, vision, live status, and a chat that drives real work.
- Maestro — guardrails, memory & cost telemetry bolted onto coding agents.
- Intel Arc stack — local LLM serving on Arc B60 / Battlemage. budget silicon, real throughput.
| repo | what it does |
|---|---|
| intel-arc-llm-stack | run any LLM locally on Intel Arc. the stack i wish existed. |
| muse-maestro | guardrails, memory & cost telemetry for Meta's Muse Code agent. |
| sentinel-code | multi-provider AI coding agent with smart model routing. |
| intel-arc-llm-benchmarks | real, production-tested inference numbers on the Arc B60. |
| msp-tier3-toolkit | battle-tested PowerShell from real MSP helpdesk ops. |
see it running — self-hosted AI you can actually touch: dirtysouthalpha.com → an interactive Neuralis brain, the AURORA-X control dashboard, and a Sentinel Prime autonomous run.
Brandon Goolsby (he/him) — Senior AI Engineer · 20+ years in cybersecurity. Security+ · ISC2 CC · CCNA-level · SonicWall SNSA. i build autonomous AI systems and open-source the parts worth sharing.
self-hosted AI · security-first · always innovating — every repo shares one identity (brand kit)
Dirty South Alpha™ · © 2026 · dirtysouthalpha.com · self-hosted AI · security-first