About this Project

In this project, I used AWS services like Bedrock, to create a cost-effective LLM for players of VALORANT which can provide information such as player information, past games, historical data etc.

Inspiration

This project was inspired me as a first-year Artificial Intelligence Bachelors student when I came across the Amazon Generative AI orchestration tool of Bedrock. It served as an inspiration to be practical in the emerging field of Gen AI and work on solutions in AI in tandem with core Machine Learning concepts to be effective in creating overall solutions.

New Concepts

It was my first time opening an AWS account and being introduced to AWS services like S3, Bedrock, SageMaker and concepts like Knowledge Bases within Bedrock. I learnt the entire process of retrieval of data to the deployment of it into a LLM and also how costly an LLM can get. It was also my first time coding with boto3.

How the Project is Laid Out

The project makes use of a portion of the publicly available s3 bucket data for the entire project. It is to create a proof of concept for the process of accessing s3 data to create Knowledge Bases in Bedrock. The process involves the following steps:

Using a python script, I have

• Extracted the data from the s3 bucket using a Lambda function
• Converted the gzip files to raw data within the same function
• Reuploaded back onto the s3 bucket ready for the Knowledge Base

Challenges Faced

Time. Learning all concepts by reading documentation, watching tutorials and following expert advice is time-consuming as a first-time learner. Using a lot of trial and error in many places consumed enormous amounts of time

What's Next?

Improve the project to include all data. Make the data optimal for direct insertion into the Knowledge Base and also monitor costs more effectively as it is a concern for businesses to optimise costs.

Built With

• amazon-web-services
• bedrock
• boto3
• knowledge-bases
• lambda
• python
• s3