Los valienties

LLM-powered digital assistant with Amazon Bedrock to build teams and answer questions about VALORANT Esports players

Hackathon Project Summary: Data Analysis and AWS Architecture

This project is a sumbmission of the VCT 2024 hackathon, we where required to build a LLM-powered digital assistant with Amazon Bedrock to build teams and answer questions about VALORANT Esports players.

Final Streamlit app

Initial Planning and Data Analysis

In the beginning, the hackathon project focused on understanding the vast amount of data available (*101 GB compressed). Using the jq command, patterns and features were extracted to build a knowledge base for an LLM FM.

Data Processing and Flattening

JSON Data: The data from the S3 bucket was processed using the jq tool to extract relevant events. The goal was to create a database schema to evaluate player performance through features like damage dealt, players killed, and assists.

PowerShell and Python Tools:

  • PowerShell Script: Used to filter and summarize large VAL data sets.
  • Python Script: Developed to process JSON files using user inputs and convert them to CSV for integration with AWS.

Text Extraction: Textual data was extracted from XML files using Beautiful Soup and other libraries to support the Amazon Kendra knowledge base.

AWS Architecture

Following the tutorial “Develop advanced generative AI chat-based assistants by using RAG and ReAct prompting,” the project utilized various AWS services like Amazon S3, AWS Glue, and Amazon RDS. The solution, named Los Valientes, leverages Amazon Bedrock and Streamlit for team building and providing insights into Valorant Esports players, enhancing the user experience with an LLM-backed chat-based assistant.