LLM powered digital assistant for VCT Hackathon

Empowering teams with an AI-driven digital assistant for building teams and getting insights on players.

Comment

About the Project

As the VALORANT esports industry continues to grow, team managers and coaches are increasingly relying on data to make informed decisions about player recruitment and team composition. However, manually analyzing performance metrics can be time-consuming and prone to human error. You are hired as a data scientist on a new VALORANT esports team and have been tasked by the team’s general manager to support the scouting and recruitment process. The idea was to develop an LLM-powered digital assistant that could automatically analyze player performance data and recommend the best team compositions based on specific roles and synergy while answering various questions about VALORANT esports players.

Inspiration

I’ve always been fascinated by how data science could be incorporated in the esports industry and competitive games like VALORANT. Watching professional matches, I realized that the composition of a team plays a crucial role in success—it's not just about having the best individual players, but also ensuring that they complement each other within a strategic framework. This project was an opportunity to combine my passion for gaming and my skills in AI and cloud development to create something impactful in the esports world.

What I Learned

This project was a deep dive into the world of AI-driven decision-making and cloud architecture. I gained a deeper understanding of AWS services, particularly how to integrate large language models (LLMs) with AWS Bedrock Agents to solve complex, real-world problems. Additionally, I learned how important it is to have a clean and well-structured knowledge base—in this case, the player performance data stored in Amazon S3. The quality of the output from the model was directly related to the quality of the data it was trained on and the response the LLM gave.

Conclusion

Building this digital assistant was a rewarding experience that combined my interests in gaming, AI, and cloud development. The assistant successfully streamlines the scouting and recruitment process for VALORANT Esports by making data-driven decisions about player selection and team composition. Through AWS Bedrock and thoughtful prompt engineering, the model was able to analyze player performance metrics and generate teams that not only had strong individual players but also worked well together as a unit. This project has paved the way for future innovations in data-driven player recruitment and optimization in the esports industry.

Testing Instructions

It is recommended to just test the application on the url provided.

  1. Frontend: The frontend is hosted on AWS Amplify and can be accessed via the provided URL.
  2. Backend: The Django backend is hosted on AWS EC2 and can be accessed through API endpoints that interact with AWS Bedrock for player analysis.
  3. Data: The model uses the most recent 90 days of VALORANT player data from Amazon S3 for team composition recommendations.
  4. Test the Assistant: Input different player statistics and request a team composition to see how the AI evaluates players based on predefined roles.
Share this project:

Updates