Product
Product Vision
Product Hardware Vision

Product hardware vision

  • Restructure contents of this file into more appropriate files / folders (contents cover not just hardware, but software too)
  • I need separate files to keep the vision for each particular system (CI-platform, CD-platform, etc.)

Enterprise-helper as a solution for AI-enabled self-hosting of apps, libs, services, models, AI.

  • Consider designing and developing Enterprise-helper as an end-to-end solution for AI-enabled & self-healing, self-hosting of apps & services.
  • This is based on my observations when searching for a cloud provider of managed k8ts. It was quite expensive to host it.
  • Why not have your own k8ts and physical server? Managed by AI, with which you can interact with natural language. AI that handles everything, each aspect of it, so that you don't have to deal with it on your own. With physical servers & required hardware and software, pre-setup for Enterprise needs, all open-source. Configured into packages, tailored to different team-sizes / goals. From individual, small teams to Enterprise.

The point is, that if such AI will be able to manage a physical server end-to-end, self-heal and reduce the manual maintenance, required for it, then Enterprises themselves might be interested in it.

Next step in this logic. It would be logical for producers of hardware to include AI into their hardware systems, so that it is shipped with AI onboard right away. Enabling to interact with everything in their hardware through a standardized AI-interface and protocol, with just a natural language. And to achieve anything done previously with older interfaces (what are those? Cli-commands? Else?) in a much simpler way.

Should I consider building (or assembling from existing products (e.g. compute blade)) a compute cluster with task-specific hardware for:

Imagine scalable production-ready full-cycle AI-app-development hardware-stack as a commodity. (At such a cheap affordable price, that anybody can buy it. Any individual. And scale it up to meet enterprise compute volumes, if needed).

CI-server

devnw/atomizer (opens in a new tab) - for parallel execution

Imagine thousands of machines across multiple cloud instances and data centers executing simultaneous processing for you with minimal deployment effort. Enter Atomizer - a Go library that facilitates massively parallel and distributed computing. Fashioned after Go itself using similar CSP paradigms.

CI-specific hardware

Consider crafting CI-specific hardware (opens in a new tab) (& runners (opens in a new tab)?) like in Blaze (opens in a new tab) (for CI across the entire AI-dev (CI for apps, CI for LMs, etc.))

AI for CI

Study nx-cloud-ai (opens in a new tab), consider implementing features alike, e.g.:

AI for your CI

Identify and resolve task failures instantly with intelligent explanations and actionable solutions. Set your desired CI run time, and Nx Cloud will match it. Our custom AI model analyzes your previous runs, then dynamically predicts and allocates the optimal number of agents. The more you use it, the smarter it gets. Take the guesswork out of your work.

Other

Should be performant enough in order to be able to run DevOps for apps/libs, but also MLOps for LMs, Data. Ideally have an open-source solution like nx, but also cloud solution like cloud.nx.app.

CI-platform Status Page

Using hud.pytorch.org (opens in a new tab)

hud.pytorch.org (opens in a new tab)

pytorch/ci-hud (opens in a new tab)

CD Server

Local Continuous Deployment Platform like vercel (opens in a new tab) (github link) ?

CD Server Status Page

Do we need one?

AI-Server

...

AI Server Status Page

Do we need one?

Tabletop cluster (?)

Enterprise-helper Compute-blade (or any other thing it might be?) cluster can be designed in several flavors:

  1. Tabletop solution (for individual / team usage)
  2. Full-fledged rack (for enterprise)
Product VisionQuality