-
AWS GUI can be overwhelming almost like an airplane cockpit
-
Copilots assume you can be the pilot if you are not an expert, you cannot be trusted as a Pilot
-
-
We want to deploy a sentiment analysis transformers model from hugging face to Sagemaker
-
humm, looks like is a bit complicated
-
CloudPilot to the rescue
-
demo cont
Motivation
AWS is an awesome Cloud provider with lots of services continuously expanding. However for the beginner and even experts it can be overwhelming.
Wouldn't be great to have an assistant like a Copilot for this, unfortunately the Copilot paradigm assumes that you can be trusted as the Pilot and that is not always the case, specially if you are a beginner.
One problem with LLMs in enterprise applications is that not always can be trusted, they make up things hallucinate and have a mind of their own.
Enter CloudPilot, it is an LLM based Conversational AI assistant that can be trusted with your AWS account. CloudPilot can be interfaced from Slack and it leverages latest advances with Nvidia Nemo Guardrails framework which provides safety to make sure that the bot always stays on topic and performs only safe and previously blessed tasks.
CloudPilot has been provisioned with AWS blueprints that follow best practices and it not only help you with automating deployments without having to go to the GUI but also instruct you about the best practices.
NewRelic is used for monitoring and LanceDB for semantic search
What's next for CloudPilot
Going from proof of concept to productization with more blueprints and additional interfaces like VSCode plugins.
Built With
- amazon-web-services
- huggingface
- lancedb
- langchain
- new-relic
- nvidianemoguadrails
- openai
- python
- sagemaker
- sentencetransformers
- slack-api
Log in or sign up for Devpost to join the conversation.