Build assistants using your own documents and data with any of the supported large language models (LLMs), from Anthropic, OpenAI, Mistral, Meta, or Google.
Your assistants will automatically answer questions using your documents, customize their responses based on your guidance, and can be shared across your organization. Build expense, travel, HR, and other assistants in minutes.
You deploy and run Amplify GenAI yourself in your own Amazon Web Services account. You pay for the Amazon Web Services costs and the token costs for the models you use.
Case Study: Vanderbilt University currently provides unlimited usage of the platform to 1,300 pilot users. In April of 2024, the Vanderbilt cost per user was around $3/mo, including Amazon Web Services costs. The largest part of the bill was not token costs. The largest token cost was GPT-4 Turbo usage. For some organizations, per user pricing may be more cost effective. For Vanderbilt, paying for actual usage was far more cost effective. 50% of Vanderbilt users consumed less than $1/mo in token costs.
Over 300,000 people have learned to innovate with Generative AI using the courses that we have selected. Although some of the classes may have ChatGPT or OpenAI in the title, the fundamental concepts, particularly Prompt Engineering, will translate across most LLMs. Look for updated versions of these courses with Amplify GenAI soon.
Have any questions?
The platform is MIT-licensed. In order to run it, you will need to pay for Amazon Web Services costs. In addition, you will need to pay for usage of models from Anthropic, OpenAI, Meta, Google, etc.
You run the platform in your own Amazon Web Services account. Your agreement with Amazon Web Services and your agreements with the model providers determine the privacy of your data. Most organizations can sign an enterprise agreement with a provider to obtain data privacy guarantees, including not training on your data.
Yes! We are happy to review and accept pull requests. In addition, you can reach out to us using the Contact Us button.
You will need an Amazon Web Services account and experience deploying applications using Terraform and the Serverless Framework. We try to help with general or architectural questions if you Contact Us.
Jules White, Allen Karns, Karely Rodriguez, and Max Moundas at Vanderbilt University build and maintain the platform. The user interface started with the excellent MIT-licensed Chatbot-UI https://github.com/mckaywrigley/chatbot-ui. We enhanced the user interface and then built a new back-end with advanced Generative AI features and enterprise capabilities to support our needs at Vanderbilt University.
There are some great, but expensive and closed enterprise options out there. There are some fantastic non-enterprise platforms and interfaces, such as the UI we started with. However, none of the options gave us the advanced Gen AI features, enterprise security, and cost that we needed. We wanted to be able to provide the latest in Generative AI research to as many people as possible at a low cost. At Vanderbilt, we believe our approach has led to a more advanced platform at a dramatically lower cost for Vanderbilt. Currently, at Vanderbilt, our costs are ~$3/mo per user and we have access to models from the top providers. In addition, we find that our platform outperforms many or all commercial options across many tasks. There are lots of 'engines' being built (e.g., the models), but the cars around the engines have lagged behind. You can think of Amplify GenAI as the car around the engine. Amplify GenAI gives you the steering, brakes, wheels, and safety systems to use engines from many different providers. Just because you like one engine, doesn't mean you have to buy their car and only use it. Amplify GenAI gives you a choice of engine and the freedom to customize the car to your organization's needs.
We can't provide support for your deployment of Amplify GenAI, but we can help with general questions about the platform. We are happy to connect you to commercial entities that can help you set up and run the platform.
We currently support models from Anthropic and Mistral via Amazon Bedrock and OpenAI via OpenAI or Microsoft Azure.
We don't have this feature implemented yet, but it should be fairly easy to set up if you are already building your own models.
Vanderbilt University has 1,300 users in Amplify and will be releasing it enterprise-wide July, 2024.
We are building several open advisory boards that will help guide the direction of the platform. If your organization is interested in participating, please Contact Us.
