Wicked Willys Search Optimization

Inspiration

The inspiration behind this project came from how often Wicked Willy’s—a staple for NYU students—gets overlooked in local search despite being physically right where people are searching (like “bar near Bleecker Street”). It made us question: how many great businesses lose customers simply because they’re not visible online?

What it does

This system improves the discoverability of Wicked Willy’s by acting like an automated growth engine. It analyzes competitors, identifies SEO and listing gaps, and not only suggests improvements—but actually executes many of them.

We can think of visibility as a function of multiple factors:

$$ \text{Visibility Score} = \alpha \cdot \text{Reviews} + \beta \cdot \text{SEO} + \gamma \cdot \text{Citations} + \delta \cdot \text{Relevance} $$

By increasing visibility across search platforms, the system ultimately helps the business attract more customers and optimize pricing based on demand.

How we built it

We built a multi-agent system using OpenAI APIs integrated into a modular architecture developed in Cursor.

The system consists of three specialized agents:

$$ A_1 \text{: Analysis Agent} $$

$$ A_2 \text{: Action Planning Agent} $$

$$ A_3 \text{: Execution Agent} $$

The orchestration pipeline can be summarized as:

$$ \text{Query} \rightarrow A_1 \rightarrow A_2 \rightarrow A_3 \rightarrow \text{Results} $$

Where:

• A₁ (Analysis Agent)
• A₂ (Action Planning Agent)
• A₃ (Execution Agent)

Challenges we ran into

One of the biggest challenges was integration—getting multiple agents to communicate cleanly while maintaining consistent context.

We also had to balance:

• Structured outputs vs. LLM flexibility
  
• Real-world SEO logic vs. generated insights
  
• Automation vs. actions that require human intervention
  

Accomplishments that we're proud of

We’re proud that we got a full end-to-end system working, where:

• A single query triggers analysis, planning, and execution
  
• The system produces real, usable outputs (SEO content, schema markup, etc.)
  
• It goes beyond insights and actually executes optimizations
  

What we learned

• Multi-agent systems are powerful, but orchestration is critical
  
• LLMs require structure to produce reliable outputs
  

SEO performance can also be framed as an optimization problem:

$$ \max_{\text{actions}} \; f(\text{ranking}, \text{traffic}, \text{conversion}) $$

• Even partial automation can significantly improve business outcomes
  

What's next for Wicked Willy’s Search Optimization

• Add more agents (e.g., pricing optimization, social media growth)
  
• Integrate real APIs (Google Places, Yelp, SEO tools)
  
• Build domain-specific intelligence for hyper-local SEO
  
• Develop a dashboard for continuous monitoring
  

Long-term, we envision:

$$ \text{Autonomous Growth System} \approx \lim_{t \to \infty} \text{Self-Optimizing Agents}(t) $$

A system that continuously learns, adapts, and optimizes business visibility in real time.

Built With

• python