GenAI Demo Showcase

Introduction

At a time when AI services are being democratized, cloud services are not necessarily the most viable option for running such software. The fact that cloud compute power is sold on a subscription basis alone makes it expensive in the long run. Additionally, as a user of a remote cloud service, one doesn’t have full control over the machine; your private data, etc. are all controlled by the service provider.

Enter JOHNAIC, your personal AI server. You can use it to run many AI and software as a service (SaaS) applications on it and access them from anywhere while still owning your data. JOHNAIC is a cloud in a box, with a small footprint and no subscription fee whatsoever.

Ease of Use

Setup for JOHNAIC is remarkable simple, essentially plug-and-play. No special power supply is needed; it can run off a standard wall socket. It runs Debian GNU/Linux, and comes with a powerful suite of applications by default, such as

• LLAMA 3 and Ollama: LLM and LLM runtime
• OneAPI: GPU runtime library/driver
• GNU Guix: Functional package manager
• Keycloak: Identity manager
• Outline: Knowledge base solution
• ERPNext: Business software for HRMS, CRM, accounting, etc.

“Free” as in “Freedom”

JOHNAIC respects users’ freedom. All software shipped is FOSS (free and open-source). Additionally, since JOHNAIC is a one-time purchase, the user truly owns the machine once bought.

Hardware Specifications

JOHNAIC comes with following hardware:

Intel i5 12400 CPU

Intel Arc 770 16 GB GPU

64 GB RAM

1 TB SSD

1100 VA UPS (optional)

Its peak rated power draw will be 500W. Using UPS, the system will be powered on for about 15 minutes if there’s a power outage. Customer should ensure the power is back on within 15 minutes for best experience. We also recommend connecting their Wifi router to this UPS for uninterrupted remote access. If wifi router can not be moved, a UPS for wifi is also sold separately.

About the presenter

Vedanth Padmaraman is a 3rd-year student of Computer Science and Engineering, and is currently working on compiler design as an intern at Von Neumann AI.