Project Alpha
Custom AI platform
Project Alpha is a secured, custom AI platform that leverages several different Large Language Models (LLM) and Retrieval Augmented Generation (RAG).
At this time, it is a proof of concept designed to demonstrate how custom, secured, and offline AI can support a variety of specific needs.
Welcome to RBC Lab's Project Alpha!
Project Alpha is my first AI proof of concept. I'm happy to share it with a small, trusted community, as well as potential investors. See How to Access below.
How does it work?
Using Open WebUI and Ollama, over the past months I've created a home AI server. See my blog article for insight into that journey. The AI server leverages multiple AI models I've personally customized, and even RAG for higher quality answers. What's even better is that all of it operates in a local, containerized environment that I'm able to take offline while preserving full functionality! In short: custom and private AI available both locally and without the internet. Having secured access through NginX, Cloudflare, and a secured custom domain, I'm proud to announce I can provide a small community of users with secured trial access!
What to expect.
This GPT-like environment provides customized chats through text or voice, and even has an interactive mode. Certain users can upload documents, gain updated information past a model's training date through the Web search, generate custom pictures through the Image generation capability, and even leverage a Code Interpreter.
In the main chat area, you can select which model you want, such as the base models and custom models. You can add another model to the chat, and even ask another model, mid chat, to provide input. Certain users can select from at least three base models, and up to multiple custom AI models that I've personally tuned. Some models can even use RAG for a very custom experience.
Why a custom AI platform instead of commercial AI?
Mainly loss of control and privacy. Each time a commercial generative AI platform releases a new model, those who built custom workflows or business with those models may have to then adjust. For many, privacy remains a key concern and some may argue commercial AI platforms have demonstrated the ability to satisfy. Additionally, if you or your organization has the need to operate locally, offline, and privately, then something like this might be a solution.
How to get custom AI Models in the UI.
Contact me to discuss your needs. Custom AI models consist of a base model and custom model parameters. RAG can be added if you need very specific and higher confidence responses. In this way, RAG can help reduce and potentially eliminate AI hallucinations, a common problem with generative AI. See my Blog for an article on The Entrepreneurial Need for RAG.
What this offers.
At this time, I'm offering limited, trial access, for peer development, experimentation, and demonstration of the process capability as a proof of concept. I may soon offer consultation services for those wishing to establish a secure and offline AI platform, as demonstrated by Project Alpha.
While I have been approached about providing a subscription service, that remains to be seen. I have zero capability or desire to, compete with commercial AI providers. Their capability, sheer computing power, scale, and support surpass what any private or small business can offer. At this time, Project Alpha is a private platform designed to demonstrate capability, experience, and support a small, trusted community.
How to access.
For access please navigate to the Contact page or click below. Once terms and conditions have been reviewed and an agreement struck, I'll share the URL and temporary log-in credentials.