Qlaire

• 

  Qlaire's landing page

• 

  Ads Insights Job Flow

• 

  Job status and Seed Entity display

• 

  Qloo's related entity search result and popularity scores

• 

  Ads insights, creative concepts and Ad Copy

• 

  Ads segmentation and key insights

Inspiration 🚀

I was inspired by the idea that great advertising is cultural, not just computational. Qloo’s unique ability to surface cultural correlations and affinities pushed me to imagine a tool that could empower creators and marketers to launch data-backed, culturally intelligent campaigns — without needing a full strategy team.

What it does 🎯

Qlaire helps you generate personalized ad campaigns based on a product name, platform, and theme. In just seconds, it:

• 🔍 Enriches your input with Qloo’s cultural graph:
  • Related entities
  • Popularity trends
  • Demographics
    
• 🤖 Sends the full prompt enriched with Qloo's insights and the user's inputs to a fast LLM
  
• 📦 Returns a complete, structured JSON campaign including:
  
  • 👥 Audience persona summary
    
  • 🎯 Audience segmentation rules
    
  • 📝 Ad copy (headlines + descriptions)
    
  • 🎨 Creative concepts (image/video ideas)
    
  • ⚙️ Campaign settings (objective, placements, budget, A/B tests)
    
  • 💡 Key insights to guide strategy
    

How I built it 🛠️

• Backend: Go (Golang) with clean architecture and an async job lifecycle management
  
• LLM Integration: Groq’s ultra-fast inference API, configured to use Llama 4 model (that can be iterated easily to test other LLMs)
  
• Cultural Enrichment: Qloo’s APIs for demographic & entity insights
  
• Frontend: React
• Hosting: Render.com
  

Data Flow:

1. POST /ads-insights → queue job
   
2. Generate a seed concept an LLM
3. Fetch Qloo data
4. Build prompt with insights
5. Generate Ads campaign with an LLM
   
6. Update in-memory results and & serve results via GET /ads-insights/{job_id}
   

Challenges I ran into 🧩

• Prompt engineering was tricky: LLMs tend to break schema easily — I needed strict formatting and cleaning. At the end, including some retries (behind the scenes) in all LLMs interactions helped to deliver a more robust solution.
• Building a concurrency-safe job queue with retries & timeouts in monolithic hackathon service
• Handling the flexible LLM interfaces to interact with Qloo's REST API.
• Crafting a smooth UX around async jobs and handling incomplete LLM responses
  
• Designing a logo & brand identity that feels as smart and sleek as the product I imagined initially

Accomplishments I am proud of 🏆

• Fully working prototype in very short time with structured JSON output, clean frontend, and live deployment
  
• Seamless integration of Qloo insights + Groq LLMs in one cohesive UX
  
• A creative tool that could save hours of campaign strategy work and can be easily considered for A/B testing against the current Ads Campaigns
• A polished brand and interface, I tried to make Qlaire feel like a real product
  

What I learned 📚

• How to translate raw data into creative prompts that LLMs can execute
  
• The power & flexibility of Qloo’s API for cultural recommendations
  
• The importance of guardrails & retries for LLM-generated structured output
  
• That UX needs to consider edge cases (async failures, schema breaks)
  

What’s next for Qlaire 🌟

• 🌐 Add multilingual support for global campaigns
  
• 🖼️ Enable AI-generated creatives (DALL·E, SDXL)
  
• 📥 Export campaigns (PDF, Meta Ads import) directly into the platforms
• 🤖 Fine-tune a smaller LLM for offline/edge use
  
• 🎨 Evolve Qlaire into the go‑to assistant for indie devs, marketers & brands
  

Built With

• go
• groq-chat-completions-api
• postgresql
• qloo-insights-api
• react
• render.com
• tailwind-css
• vite