Inspiration
In the high-stakes world of professional esports, analytics are often trapped in the "Stat Era." Coaches and players are flooded with raw numbers—K/D ratios, damage dealt, and gold per minute—but they are starved for the context that leads to victory. We noticed that while a coach might see a player died five times, they have to manually scan hours of VODs to understand why. We were inspired to build "Echo" to bridge this gap: a tool that doesn't just record history, but reasons through it to provide the "why" behind every match-deciding moment.
What it does
Project C9 "Echo" is an AI-powered tactical reasoning engine. It acts as a 24/7 "Holistic Assistant Coach" that:
Identifies "Moneyball" Moments: Automatically filters through thousands of GRID match events to find high-leverage sequences, such as gold swings exceeding 1,000 in under 30 seconds.
Provides Tactical Reasoning: Using the persona of legendary Cloud9 Coach Inero, it analyzes event telemetry to identify specific tactical errors like spacing lapses, utility mismanagement, or isolated deaths.
Gamifies Coachability: Features a "Tactical IQ" leaderboard that ranks players based on their strategic discipline and positioning rather than just raw mechanics.
How we built it
The project is built on a high-performance stack optimized for low-latency data processing:
Python 3.12 & FastAPI: Leveraged the latest interpreter speeds and asynchronous routing to handle high-frequency data from the GRID Platform.
Gemini 2.5 Flash: Integrated the reasoning layer to transform raw JSON telemetry into expert-level coaching insights.
GRID Data Platform: Authenticated via the x-api-key protocol to ingest real-time GraphQL match data for VALORANT and League of Legends.
PyCharm Professional: Our primary IDE, where we utilized JetBrains Junie to guide our architectural decisions and maintain a modular, production-ready codebase.
Vercel: Deployed as a full-stack serverless application, ensuring the dashboard is accessible across devices.
Challenges we ran into
The biggest technical hurdle was Data Normalization. GRID provides incredibly detailed data, but the "shape" of a kill event in VALORANT (spatial coordinates, utility usage) is vastly different from a kill in League of Legends. We built a custom normalization layer in Python to ensure our AI reasoning engine could interpret the tactical importance of a death regardless of the title. Additionally, configuring the Vercel serverless environment to correctly route between a static HTML frontend and a FastAPI backend required precise rewrite rules.
Accomplishments that we're proud of
We are incredibly proud of the "Reasoning Latency" we achieved. By using Gemini 2.5 Flash, we are able to provide a complete tactical review of a complex 30-second play in under 3 seconds. We also successfully implemented a dark-mode optimized "Coach Terminal" UI that feels like a professional tactical tool rather than a generic stat site.
What we learned
This project taught us the power of Agentic AI in niche domains. We learned that an AI model is only as good as the context it's given; by feeding the model specific "Moneyball" metrics like gold deltas and spatial coordinates, we could prevent generic advice and get the AI to provide professional-grade critiques on spacing and utility trade-offs.
What's next for Project-C9-Echo
The vision for Echo is to move from post-match analysis to Live Tactical Overlays. We plan to integrate direct video-timestamping, where the AI's critique is linked to a specific frame in a VOD. We also want to expand the "Tactical IQ" system into a full recruitment tool for orgs like Cloud9 to identify "undervalued" players who have high strategic discipline even if their raw stats don't scream "MVP."
Log in or sign up for Devpost to join the conversation.