Every year, the Super Bowl attracts 100+ million viewers in the U.S. alone. It’s one of the rare events where sports fans, casual viewers, and even non-sports watchers all tune in together. 

That attention is incredibly valuable yet incredibly expensive. A single 30-second Super Bowl ad costs millions of dollars, and brands often spend even more on production, celebrities, and cross-platform campaigns. But small businesses are shut out of that moment. Yet their customers watch the same game, scroll the same feeds, and react to the same moments.

Madison Avenue has 6 months and 500 people to plan for this game. You have 5 hours, your team, and your imagination to beat them.

Build an AI-powered tool that helps medium sized businesses capitalize on Super Bowl buzz in real time. 

This is not about producing a Super Bowl ad. It’s about real-time reaction, smart automation, and real business impact.

Your AI system should help a small business:

• React quickly to what’s happening during the Super Bowl

• Generate timely, relevant marketing content

• Launch or suggest promotions tied to live moments

• Turn cultural buzz into customer engagement or sales

Think in terms of: “Something just happened — what should this business do right now?”

Your project must use real data from live APIs OR process visual data for actionable insights.

Bonus points for both.

Your project must be publicly accessible and functional during judging.

• Deploy your application to a live environment (not just localhost)

• Provide a working URL or access method

• Demonstrate the full flow live: Real Data → AI Processing → Business Output

1. Real-World Data Integration

Did the team successfully use live external data?

2. Live Deployment

Is the product working in a real, deployed environment?

3. Latency

How long elapsed between the TV "moment" and the AI's "response"?  (Target: < 45 sec.)

4. Business Impact

Would this realistically help a small business grow engagement or revenue?

5. Smart Use of AI

Is AI meaningfully powering content, decisions, or automation?

6. Usability

Could a busy small business owner actually use this without 2 million steps?

7. Demo Quality

Does the team clearly show: Live Data → AI Output → Business Action

Ingestion (Multimodal): Gemini 1.5 Flash (or 2.0 Flash). Feed live video frames/audio chunks directly to the model to detect events without stitching disparate APIs.

Reasoning (Speed): DeepSeek R1 via Groq. Use this for sub-second strategic decisions ("It's a fumble -> Generate a 'Fumble-Proof' insurance ad").

Orchestration (Control): LangGraph. Build a state machine that maintains "Game Context" and only triggers expensive generation steps when specific confidence thresholds are met.

Generation (Visuals): Flux 1.1 Pro (via Fal.ai). The current standard for speed + text rendering capabilities (crucial for legible discount codes/slogans).

Frontend (The "Wow" Factor): Vercel AI SDK. Use the Data Stream Protocol to visualize the AI's "thought process" live on the dashboard, proving it's reacting in real-time.

Pro Tip for the Demo: Use a YouTube clip of a past Super Bowl and feed that into your system live during the pitch. Don't rely on an actual live broadcast during your demo (nothing might happen!). Show it processing "The Catch" or a "Fumble" to prove it works instantly.

• Creativity
  
• Technically Sound
  
• Commercializability
  
• Presentation