Blue FC AI Studio

Cultural intelligence meets sports passion - AI that personalizes merchandise and content for individual fan preferences

Comment

BlueFC AI Studio - DevPost Project Story

Inspiration

The inspiration came from two worlds colliding - my professional life as a data scientist and AI engineer, and my passion as a lifelong Chelsea FC supporter. I've always dreamed of creating something meaningful for my favorite club, and when I discovered Qloo's cultural intelligence capabilities, I realized we could solve a massive problem in sports marketing.

Current fan engagement is broken - millions of diverse, global Chelsea fans get identical generic experiences. A tech-savvy supporter in Singapore sees the same products as a traditional fan in London. I knew there had to be a better way to understand what fans truly want based on their cultural preferences, not just demographics.

Mixing my technical expertise with my Chelsea obsession felt like the perfect opportunity to build something that could genuinely impact how sports brands connect with their fans.

What it does

BlueFC AI Studio transforms sports fan engagement through three core AI-powered capabilities:

1. Smart Product Recommendations: Users simply ask in natural language like "What products would appeal to young Chelsea fans in Toronto?" Our Google ADK agents parse the query, extract cultural signals, and use our custom Qloo Python wrapper to get deep cultural insights. The result: merchandise recommendations with actual cultural compatibility scores.

2. AI Product Customization: Upload any Chelsea merchandise image, select a target demographic, and watch our AI redesign it for that specific cultural context. A standard jersey becomes a minimalist tech-inspired design for Singapore professionals or a heritage-focused variant for traditional London fans.

3. Personalized Video Content: Users input their profile and interests, our system analyzes their cultural preferences through Qloo API, then generates custom Chelsea videos woven with their lifestyle, location, and interests. Each video includes culturally-targeted ads from brands that match their taste profile - turning personalization into revenue.

The magic happens through our Google ADK agent orchestration, which seamlessly converts any English query into Qloo cultural insights, then transforms those insights into personalized experiences.

How we built it

Architecture:

  • Google ADK: Agent orchestration and workflow management
  • Custom Qloo Python Wrapper: Built from scratch to interface with Qloo's cultural intelligence API
  • Vertex AI Gemini: For natural language processing and cultural analysis
  • Vertex AI Imagen: For AI-powered product customization and visual content
  • Supabase: Database for product catalogs and user preferences

Development Process:

  1. First, I spent considerable time understanding Qloo's API structure and building a comprehensive Python wrapper covering most endpoints
  2. Integrated this wrapper as tools within Google ADK agents, enabling natural language to cultural insight conversion
  3. Built the product recommendation engine using Qloo's audience and entity APIs
  4. Developed the AI customization pipeline using Imagen models guided by cultural preferences
  5. Created the personalized video generation system combining user profiles with cultural insights
  6. Deployed everything as agents in Vertex AI's agent engine for seamless orchestration

Challenges we ran into

Qloo API Mastery: Understanding Qloo's cultural intelligence API was initially overwhelming. I went in circles trying to grasp how entities, audiences, and cultural signals interconnected. Building a robust Python wrapper that could handle all the nuances took significant iteration and testing.

Agent Deployment Complexity: This was my first time deploying agents to Vertex AI's agent engine. Understanding how to properly orchestrate multiple AI services, handle state management, and ensure smooth data flow between components was a steep learning curve.

Video Content Generation: Creating a system that could generate genuinely personalized video content at scale while maintaining quality and cultural relevance required careful prompt engineering and multiple model orchestration.

Time Constraints: Balancing my day job as a data scientist with late-night development sessions meant every hour counted. Some ambitious features had to be scoped down to meet the hackathon timeline.

VIdeo Content Generation deployment online: The video content development is done by a complex process (To have more control than using Veo and lower cost). FIrst we create scenes, then images for the scenes with imagen, audio script for the scenes later converted to audio with google TTS. Then they are compiled automatically into a video by pymovie. This whole process take about 8 minutes. The challenge with streamlit app deployed on cloud run with normal resources (max 4 instances) was that it did soft refresh every 5-6 mins. Due to this the session connection to content agent deployed is lost midway. And hence the video content tool won't work compleletely with the streamlit app deployed online to full, on videos longer than 15 seconds. Found this the long way and had to drain lot of GCP credit and time to understand this. However, as you can see in the demo the whole process works smoothly in ADK web UI and local process.

Accomplishments that we're proud of

For My Favorite Club: I'm incredibly proud to have built something meaningful for Chelsea FC. Turning my passion into a product that could genuinely help my club generate additional revenue and improve fan experiences feels amazing.

Technical Innovation: Successfully created a comprehensive Qloo Python wrapper that makes cultural intelligence accessible to any developer. The seamless integration of this wrapper as ADK agent tools demonstrates how natural language queries can unlock powerful cultural insights.

Real Business Impact: The revenue projections aren't just theoretical - with Chelsea's actual £100M merchandise revenue with 40M+ from online, even a 5-10% improvement from our personalization represents £2- 4M additional annual revenue with minimal infrastructure costs.

Privacy-First Approach: Proud to have built a solution that delivers deep personalization without collecting personal data, using Qloo's aggregated cultural intelligence instead.

End-to-End Solution: From query understanding to product customization to video generation - we built a complete platform that showcases the full potential of cultural AI in sports marketing.

What we learned

Qloo's True Value: Working deeply with Qloo's API revealed how powerful cultural intelligence really is. In an era of increasing privacy scrutiny, Qloo offers rich personalization insights without touching PII. We get everything we need to understand fan preferences through cultural signals rather than personal data collection.

Agent Orchestration: Learned how to effectively use Google ADK to coordinate multiple AI services, manage complex workflows, and create seamless user experiences across different AI capabilities.

Cultural Personalization Impact: Discovered that cultural taste-driven personalization dramatically outperforms demographic-based approaches. When you understand someone's music preferences, fashion taste, and lifestyle choices, you can predict their product preferences with remarkable accuracy.

Market Opportunity: Realized we're potentially first movers in applying cultural intelligence to sports marketing at this scale. The competitive advantage could be substantial for early adopters.

What's next for BlueFC AI Studio

Todd Boehly Integration: My ultimate dream is for Chelsea's ownership to see the potential here and let us build this into a full-scale product for the club. The revenue impact and competitive advantage could be transformational.

American Market Expansion: One major opportunity is creating content specifically for American families unfamiliar with football. Qloo could help us understand exactly what resonates with different American cultural segments to build new Chelsea fans from the ground up.

Advanced Features: Several ideas didn't make the hackathon timeline:

  • Stadium experience personalization using location and cultural data
  • Cross-sport cultural intelligence expansion
  • Creating Fan Zone for 5th stand app, by bringing fans with similar cultutal taste across globe together. This can be live game, virtual fanzones where they can chat with similar taste fans across globe. This let's us understand the pulse of fans in real time instead of getting noisy and emotional info always from twitter etc.
  • Creating more personzalized contents: Custom messages with cultural taste added from players to premium blue fans, digital products like comics with stories embedding fan and players into a theme with users taste insights from Qloo controlling the narrative Platform Scaling: The infrastructure is designed to handle any sports team or entertainment brand. The same cultural intelligence that works for Chelsea can transform fan engagement across all sports.

Global Expansion: With Qloo's cultural intelligence covering diverse global markets, we can help Chelsea (and other brands) create genuinely localized experiences for fans worldwide.

The future of brand engagement is cultural intelligence, and we're positioned to lead that transformation - starting with the best football club in the world. 💙

Built With

Share this project:

Updates