Inspiration Our love for Valorant and passion for following the game’s competitive scene inspired this project. As engineers, we saw an opportunity to combine our interest in both gaming and technology by creating a meaningful tool for the Valorant community.
What it Does The project generates fantasy teams for VCT events based on historical game data. It analyzes player statistics to build optimized teams for different roles, helping users engage with the competitive scene in an interactive way.
How We Built It We utilized AWS agents to build the backend. Instead of relying on heavier knowledge bases, we deployed five lightweight agents to pull player statistics from a database. These agents work in parallel to collect the necessary data, and the results are processed by a summarization agent with Valorant-specific context. This agent evaluates player performance based on statistics, using the role-specific insights we designed to maximize fantasy team performance.
Challenges One of the biggest challenges we encountered was managing rate limits on our backend processes, which slowed down data collection and analysis. However, through careful optimizations and adjustments to our workflow, we were able to navigate these limitations and maintain system efficiency. Accomplishments We are proud that the system functions as intended and successfully generates competitive fantasy teams. A key accomplishment was devising our own algorithm for evaluating players based on their roles, which enhanced our understanding of both the game and backend development. Additionally, we deepened our expertise in AWS Bedrock and agent-based architecture.
What We Learned This project provided hands-on experience in working with rate limits and optimizing agent workflows. We also learned how to design lightweight systems that balance efficiency with functionality, an essential lesson for real-world applications. The project expanded our knowledge of cloud infrastructure and gave us a clearer understanding of how to use AWS agents effectively.
Future Steps Looking ahead, we aim to reduce the impact of rate limits by implementing a more scalable solution. We plan to create a distributed agent flow where each agent functions as a node, enabling faster parallel processing. Additionally, we want to integrate a summarization model on the frontend to give users immediate feedback, streamlining the process and enhancing user engagement.
Built With
- amazon-web-services
- bedock
- flask-api
- lambda
- postgresql
- rds
Log in or sign up for Devpost to join the conversation.