Rift Rewind — DuoCoach AI

🌌 Inspiration

Every League player loves the end-of-year recap — that nostalgic moment where you see how much you’ve grown (or tilted) throughout the season. But existing tools like op.gg or Mobalytics focus on raw stats — not personal storytelling or coaching. We wanted to bring Spotify Wrapped energy to League of Legends, combining data-driven insights with Generative AI personality powered by AWS Bedrock.

🧠 What it does

Rift Rewind — DuoCoach AI transforms your full-year League of Legends match history into a personalized, shareable recap.

Using the Riot Games API and AWS AI services, it:

Analyzes hundreds of matches to detect your playstyle trends, growth, and clutch patterns. Generates Roast & Boast recaps with witty, AI-generated commentary. Visualizes your Consistency, Clutch, and Recovery scores. Highlights your Hidden Gems — underrated champions with high win rates. Produces shareable summary cards for social platforms, giving every player their “League Wrapped” moment.

🏗️ How we built it

We combined the Riot API for match data with an **end-to-end AWS analytics + AI stack:

Architecture Overview:

1. AWS Lambda — Fetches match history from the Riot API and saves JSON to Amazon S3.
2. Amazon SageMaker Processing — Cleans, flattens, and transforms match data into Parquet for querying.
3. Amazon S3 + Athena — Stores curated data and computes performance metrics.
4. Amazon Bedrock (Claude / Titan) — Generates natural-language summaries (“Roast & Boast” and coaching insights).
5. Strands Agent SDK — Orchestrates retrieval (RAG), summary generation, and card creation.
6. Next.js App Router + Tailwind + framer-motion— Provides a smooth, animated web dashboard hosted on AWS Amplify.

🧩 Challenges we ran into

Rate limits on Riot’s Match API — required batch Lambda scheduling. Data normalization — each champion and lane has unique metrics; aligning these into one schema took iteration. Generative cost optimization — fine-tuning Bedrock inference to stay within token limits. Promise rendering errors in Next.js— resolved by separating Server Actions from Client Components cleanly. Time— integrating SageMaker + Bedrock + frontend polish in under a week was tough!

🏆 Accomplishments that we’re proud of

Built a fully working AI recap MVP using AWS-native tools. Designed a human-feeling AI coach that provides personality and empathy, not just stats. Created a clean, animated UI that feels like a real "League Wrapped” experience. Passed all runtime test assertions for analytics accuracy (KDA, Consistency, Clutch, Hidden Gems). Maintained cost-efficient architecture using small AWS services and on-demand invocations.

📚 What we learned

How to orchestrate a complete RAG + Agent workflow with Bedrock Knowledge Bases. The importance of client-server separation to prevent rendering Promises in React Server Components. effective data storytelling — turning numbers into narratives players care about. Building delightful UX on top of technical AI systems can make data feel alive. AWS Bedrock’s APIs for Knowledge Base + model orchestration are fast and flexible for generative recaps.

🚀 What's next for Rift Rewind — DuoCoach AI

🔗 Real-time Riot integration — authenticate users and fetch live ranked data. 🤝 Duo Synergy Mode — compare performance with your most-played duo. 🧠 Bedrock Fine-Tuning — train the Roast & Boast agent to reflect personality styles (e.g. mentor vs. trash talker). 📱 Social Sharing Cards — one-click image export for Discord, Twitter, and Reddit. 🌍 Multi-Game Expansion — extend to Valorant, TFT, and Wild Rift. 💾 Deploy full AWS pipeline with Step Functions and cost-aware Bedrock inference scaling.

Built With

• athena
• lambda
• next.js
• s3
• sagemaker
• strands
• tailwind
• typescript