TeamLens - Devpost Submission

Inspiration

I wanted to democratize esports intelligence. What if anyone could generate a professional-quality scouting report in seconds instead of hours?

What it does

TeamLens is an AI-powered scouting tool for VALORANT and League of Legends. Users simply:

1. Select a game
2. Search for any professional team
3. Click "Generate Report"

In seconds, they get a comprehensive scouting report including:

• Team Overview - Playstyle, strengths, recent form
• Player Analysis - Individual stats, roles, and tendencies for each roster member
• Strategy Breakdown - Tactical patterns and team compositions
• Counter-Recommendations - AI-generated suggestions for beating the team
• Match History - Recent series results and tournament performance

No sign-up required. Just instant esports intelligence.

How I built it

The stack combines real-time esports data with fast AI inference:

Frontend:

• Next.js 16 with App Router for the UI
• React 19 with TypeScript
• shadcn/ui and Tailwind CSS for a clean, responsive design
• Vitest for testing

Data Layer:

• GRID Esports API provides real-time match data, player statistics, and roster information
• Custom service layer aggregates data from multiple GRID endpoints

AI Analysis:

• Cerebras AI with Llama 4 Scout model generates the analysis
• Structured prompts ensure consistent, high-quality output
• The model receives match context and produces insights in a predictable format

Architecture:

• API routes handle data fetching and AI generation
• Session-based report storage (no database needed)
• Responsive single-page design - everything happens on the homepage

Challenges I ran into

1. GRID API Complexity The GRID API is powerful but complex. I had to make multiple calls to different endpoints (teams, series, players) and aggregate the data into a coherent context for the AI. Getting the right data shape took significant iteration.

2. AI Output Consistency Early versions produced inconsistent analysis - sometimes too verbose, sometimes missing key sections. I refined the prompts extensively to get reliable, structured output that covers all the important aspects of a scouting report.

3. Handling Missing Data Not all teams have complete data. Some players lack statistics, some series are missing details. I built fallback handling throughout to gracefully degrade when data is incomplete rather than failing entirely.

4. UI/UX Balance Scouting reports contain a lot of information. Finding the right balance between comprehensive data and clean presentation was challenging. The tabbed interface with collapsible sections solved this - users can drill into details without being overwhelmed.

Accomplishments that I'm proud of

Speed: Reports generate in seconds. What used to take hours of manual work happens almost instantly.

Quality: The AI-generated insights are genuinely useful - not generic fluff. The model picks up on real patterns in the data.

Accessibility: No sign-up, no paywall, no complexity. Anyone can use TeamLens immediately.

Polish: The UI feels professional. The vibrant esports-inspired theme, smooth animations, and thoughtful interactions make it enjoyable to use.

Testing: 51 passing tests including property-based tests for core functionality. The codebase is solid.

What I learned

Prompt Engineering Matters: The difference between mediocre and excellent AI output often comes down to how you structure the prompt. Giving the model clear context, explicit formatting requirements, and examples dramatically improved results.

API Design is Hard: Aggregating data from multiple sources into a coherent shape requires careful thought. I learned to design the data flow before writing code.

Less is More: Early versions had more features - user accounts, saved reports, comparison mode. Stripping it down to the core experience made the product better, not worse.

Cerebras is Fast: The inference speed is remarkable. It enables the "instant" feel that makes TeamLens compelling.

What's next for TeamLens

Short-term:

• Team comparison mode (side-by-side analysis of two teams)
• PDF export for offline sharing
• More detailed player statistics
• Historical trend analysis

Medium-term:

• Additional games (CS2, Dota 2, Rocket League)
• Real-time match tracking during live games
• API access for third-party integrations
• Mobile app

Long-term:

• Predictive analytics (match outcome predictions)
• Custom report templates
• Integration with coaching tools
• White-label solution for esports organizations

The vision is to make TeamLens the go-to platform for esports intelligence - useful for everyone from casual fans to professional analysts.

---

Built with: Next.js, React, TypeScript, Tailwind CSS, shadcn/ui, GRID Esports API, Cerebras AI

Try it: https://team-lens-gg.vercel.app/

Built With

• cerebras
• grid
• jetbrain
• junie
• nextjs
• react