This site documents an end-to-end engineering journey that bridges the gap between foundational infrastructure and modern AI workloads. It covers the complete stack: provisioning KVM-based virtualization, architecting storage tailored to the task (from robust Ceph clusters to high-speed NFS), and deploying essential core services such as PostgreSQL, MariaDB, and Gerrit.
Beyond the basics, the focus extends to advanced capabilities—enabling GPU passthrough for AI, building custom cloud images, and designing full-scale software solutions. Each post captures the exact steps, configurations, and insights gathered along the way, creating a transparent, reproducible reference for building a production-grade environment at home.
Infrastructure overview
Management Server for Home Lab
Install KVM and build a custom cloud image
Orchestrating VMs using scripts and templates
Ubuntu 22.04 Repository Mirror
WordPress for documenting and sharing
PCI Passthrough NVIDIA Tesla L4/P4 on Ubuntu 22.04
Installing Ceph Reef (v18) Installation and Initial Configuration
Preparing a Custom Cloud Image for AI Workloads
NFS Server for PoCs (Ubuntu 22.04)
AI-Assisted Purchase Order Processing: A Practical Proof of Concept
– Part 1: Solution Approach and System Architecture
– Part 2: Technology Stack, Services, and Model Choices
– Part 3: Document Ingestion, AI inference, and Human-Driven Workflow
– Part 4: Human-in-the-Loop Review and Retraining Workflow
– Part 5: Implementation – Inference Service – Part 1
– Part 6: Accuracy tuning, confidence scoring, and failure modes.