QuantRift

Inspiration

We wanted to build something that goes beyond surface stats like KDA or win rate. Most League of Legends tools explain what happened in a match, but not why. QuantRift was inspired by the idea that AI can bridge this gap—analyzing real match data to give players context-aware, actionable feedback.

What it does

QuantRift is an AI-powered analytics and coaching platform for League of Legends. It analyzes gameplay across nine modules:

Performance Analysis
Champion Recommendation
Timeline Deep Dive
Role Specialization
Version Trends
Friend Comparison
Champion Mastery
Build Optimization
Annual Summary

Each module interprets match data from the Riot ecosystem and generates insights with AWS Bedrock Claude models, providing personalized feedback and improvement paths.

How we built it

Frontend: Next.js 15 + React 19, TypeScript, Tailwind CSS, Framer Motion
Backend: FastAPI (Python 3.11) + Uvicorn, asynchronous data ingestion
AI Layer: AWS Bedrock Claude 3.5 Haiku and Claude 4.5 Sonnet for natural-language reasoning
Data Sources: Riot Games API, OP.GG MCP, Data Dragon, Community Dragon
Infra: Docker multi-stage builds and Docker Compose orchestration for modular services

Challenges we ran into

Handling Riot API rate limits and large asynchronous data requests
Coordinating real-time updates across nine independent AI modules
Optimizing model latency for live feedback without degrading UX
Managing data consistency between the analytics engine and the front-end interface

Accomplishments that we're proud of

Built a fully modular, production-grade AI analytics platform within hackathon constraints
Integrated real-time Riot API data with Bedrock AI reasoning for the first time
Designed a scalable data pipeline and clean modular architecture
Delivered a modern, responsive UI with rich animations and visualization

What we learned

We learned how to structure end-to-end AI-driven analytics pipelines, how to optimize async data flow for low latency, and how to align LLM feedback with quantitative data. We also deepened our understanding of prompt design, rate-limit handling, and scalable backend orchestration.