← Back to Blog

Personalized Fashion Adv

Personalized Fashion Adv

Blog

🚀 From Graduation Thesis to Forbes Top Graduates 2025


Winning an award is always special, but seeing an academic project recognized on a national level was an amazine feeling. I am beyond proud to share that my thesis in Artificial Intelligence has been selected by Forbes Italy as one of the Top Graduates 2025!

Out of thousands of graduates from all over Italy, only 100 were chosen for this prestigious recognition.

👜 Why Fashion?


My interest in this field comes from a genuine love for fashion. I’ve always been interested in high fashion and the world built around garments, runways, and brand identity. Houses like Rick Owens, Valentino, Jacquemus, Prada, Margiela, Miu Miu don’t just make clothes; they communicate a complete visual culture, an atmosphere, and a story.

In standard e-commerce, keeping product presentations clean and objective makes total sense. The goal is to display the garment as purely as possible. However, there is a huge opportunity to push things further. That’s why I built this project:

  1. As a tool for creating advertisements or even exploring garment concepts (much like Nike’s A.I.R. project, using AI to push creative boundaries).
  2. As a feature that brands can integrate directly into their e-commerce, allowing users to see a clothing item styled and customized to their own profile on demand.

To learn more about fashion communication please have a look at these amazing resources:

⚖️ Ethics


This should always act as a support system, never as a replacement for human creativity and craftsmanship. The goal of this project isn’t to take the human element out of fashion, but to give designers and creators powerful tools to enhance their work and present it better.

More on this topic on my thesis paper.

💡 The Project: AI-Powered Personalized Advertising


By combining an individual user’s profile with a specific piece of clothing, the system leverages advanced AI architectures to create custom, tailored advertisements, drastically elevating product presentation and personalization.

As demonstrated during testing, the vast majority of users showed a strong preference for the personalized generated images, stating they would feel much more inclined to buy clothes when presented this way.

⚙️ Key Technical Features


Personalized Fashion Advertisement pipeline
Project Pipeline
  • VLM (Vision-Language Models) — Integrated to bridge visual and textual data, ensuring seamless alignment between the user profile, the garment, and the generated adv.
  • LDM (Latent Diffusion Models) — Used to create high-quality, context-aware visuals tailored to individual user preferences.
  • Performance Monitoring Dashboard — A dedicated section built into the platform to track, analyze, and optimize generation metrics and performance in real-time.

If you are interested in diving deeper into the technical details, feel free to visit the GitHub repository where you can inspect the code or read my complete PDF thesis!