Pro VCT Esports Manager

Create a Valorant team with our professional advisor! Have limited imagination? Don't worry, we have visualized Statistics and tailored Future Visions for you!

Comment

Inspiration

Imagine a Valorant site combat with mollys, smokes, mystic creatures, and grenades! How can a chatbot be more charming than that :? With Visualization and Imagination! Obviously!

What it does

Built with AWS Bedrock, Valorant Data. Deployed with EC2, Route 53. Our chatbot prepares you on your journey to create an awesome Valorant Team! We try to provide you with as much information as possible, in the most delightful manner, with the depth of strategic plans and the magic of visions.

Our chatbot can create a Valorant team to your requirements, provide detailed information about players, and tailor strategies for the newly created team. Team composition is carefully chosen with complex algorithms. Furthermore, data is carefully processed and displayed with charts and images!

How we built it

  • First: Data processing. Having hundreds of GB of data to process is not easy. We summarize the data and only keep the relevant data for production.
  • Second: AI Agent Implementation. AWS Bedrock is a fully managed AI service, that provides access to leading Foundation models and a broad range of flexible services. Processed data are uploaded into 2 S3 buckets, one for retrieval using AWS Lambda and Action Groups player data, and one for RAG (using Serverless Opensearch Service), containing general Valorant knowledge.
  • Third: Web app design: Next.js application! With AWS SDK, specifically AWS Bedrock Agent Runtime SDK. Deployed on EC2 instance, Route53 for DNS.

Challenges we ran into

  • I'm completely new to data science (well I'm good at Maths), so I have a hard time designing data pipelines with a lot of wow and wow in the process.
  • AWS Agents are new to me. But thanks to the amazing workshops, sessions, and support from the amazing people from AWS, Riot Game, and Devpost team, the process was enjoyable!
  • Choosing between a knowledge base and action groups to give knowledge to the agent is a tough choice. We managed to come up with a solution: generic information in the knowledge base, and specific algorithms in action groups.

Accomplishments that we're proud of

  • A full-stack application in 1 week! Super fun! (and a little bit frustrated : )

What we learned

  • How to leverage AWS services to build an AI Agent and integrate it with a web app.
  • How to process big data!
  • Although the application is completed, the system design wasn't the most beautiful :( A good lesson for me!

What's next for Pro VCT Esports Manager

  • INSTRUCTIONS HAVE BEEN CHANGED SINCE THE VIDEO, PLEASE TEST THE CHAT FOR NEWEST RESULT!
  • We will update the Image in the profile, as well as create more charts. We will also pre-process the chartData.
  • If possible, I would like to explore the power of native AWS data services more (Glue, Athena, ...). This time the credits didn't allow that.
  • I have many ideas about the system design, and I want to improve the current system architecture and try new AI pipeline architecture in the future.
  • I also want to allow the chatbot to answer more variations of questions (players' full history, news, ...)

Built With

Share this project:

Updates