Foundations & Hardware
Understand the physics of AI. Why VRAM and memory bandwidth dictate the rules, and how to choose between standard DDR5 PCs, Mac Studios (Unified Memory), and Multi-GPU workstations.
Zero to Hero: On-Premise AI
Sovereign, private, and high-performance architectures for LLMs. From hardware physics to enterprise deployment.
The AI Architect’s Digital Garden
Section titled “The AI Architect’s Digital Garden”The Software Stack
Discover the engines powering local models: Ollama, vLLM, how to choose the right model, and how to orchestrate a cluster using Exo or Ray.
Architecture Blueprints
Ready-to-use deployment scenarios. From a simple Dev Lab (CPU Offloading) to an Enterprise Datacenter (RoCE + Multi-GPU), with a full TCO comparison.
Agents & Assistants
Custodian agents, sovereign RAG, local interfaces (Open WebUI, AnythingLLM) and coding tools (OpenHands, Cursor CLI, Aider) — human-in-the-loop workflows, no cloud required.
Implementation
From setting up Ollama to deploying vLLM on multi-GPU: monitoring, inference security, model evaluation, and real-world migration.