sandeep In previous posts, we defined our architecture and selected our technology stack. Now is the time to build the “Brain” of the operation. This post covers the deployment of the core AI services on our GPU-enabled virtual machine. We will configure two main components: 1. Initial Server Setup First, we launch our VM based on … Read more
This document details the technical architecture, software stack, and model selection for our AI-Assisted Purchase Order Processing (POP) system. The goal of this Proof of Concept (PoC) is to demonstrate a “Self-Healing” enterprise application where human feedback directly trains the AI to correct its own mistakes without code changes. 1. Architectural Philosophy: The Unified Vision-Language … Read more
Modern Proof of Concepts (PoCs)—especially those involving microservices or Kubernetes—rely heavily on a messaging backbone. While cloud-managed queues are popular, local PoC environments often require a self-hosted RabbitMQ instance to keep costs low and latency minimal. A common architectural challenge in these environments is the “Boot Race.” If you run RabbitMQ on a separate VM, … Read more
Modern application Proof of Concepts (PoCs)—especially those involving microservices or Kubernetes—often require a shared storage layer. While complex distributed storage solutions (such as Ceph or GlusterFS) are available, they may be overkill for a lab. Sometimes, the fastest and most reliable path to a working environment is a standard NFS server. This guide covers setting … Read more
Processing purchase orders is often treated as a back-office burden, but it is the pulse of the supply chain. While traditional OCR solutions rely on rigid templates and manual data entry, modern Generative AI allows us to build a system that is flexible, autonomous, and—most importantly—self-healing. This post outlines the architecture for our AI-Assisted Purchase … Read more
This document describes the process for creating a reusable, deterministic Ubuntu 22.04 QCOW2 base image optimized for GPU-accelerated AI workloads. The resulting image is intended to be cloned and customized by automation/orchestration tooling. 1. Base OS Installation Why avoid the HWE kernel in the base image 2. System Update 3. SSH Configuration (Remote Root Access) … Read more
Platform: Ubuntu 22.04 (Jammy)Deployment model: cephadm, multi-host, mixed SSD classes 1. Install prerequisites on all hosts Install base dependencies required by cephadm and Ceph daemons: Container runtime Cephadm requires a supported OCI container runtime.Podman is the recommended and well-tested runtime for Ceph Reef on Ubuntu 22.04. 2. Kernel and networking prerequisites Enable IPv4 forwarding (required … Read more
Before deploying Ceph, it is essential to prepare the servers correctly at the hardware, BIOS, operating system, and networking layers. Many Ceph performance and stability issues originate from gaps at this stage. This post documents the preparation steps and design decisions used for my Ceph deployment. 1. BIOS Configuration All servers were first configured at … Read more
1. Why PCI Pass-through? The primary goal is to use the Tesla L4/P4 at 100% capacity within a Virtual Machine for high-speed document parsing/inference without the complexity of vGPU. 2. Host Configuration We isolate the GPU hardware, so the Guest VM can claim it exclusively. Identify the Hardware Kernel & GRUB Setup We use the … Read more
This guide walks through installing Gerrit Code Review, securing it with NGINX as a reverse proxy, and configuring HTTPS using SSL certificates.Tested with Ubuntu 22.04/24.04 and Gerrit 3.9.x. 1. Install Required Packages Run the following on the VM that will host Gerrit: 2. Create User Accounts with htpasswd NGINX uses an .htpasswd file to store … Read more