AdCraft AI

Team AdCraft AI is addressing the inefficiency and resource-intensive nature of ad creation, a challenge affecting businesses globally, particularly digital marketing teams and small businesses. Personalized ad creation is time-consuming, with marketers reporting a significant gap in access to affordable, advanced analytics and creative tools. This gap slows responses to market dynamics and customer needs, impacting performance and scalability.

To tackle this, we leverage publicly available data and cutting-edge technologies. Our approach combines Reddit data, advanced large language models (LLMs), and image generation tools. Through object detection and masking techniques, we transform user-uploaded images into polished, professional-quality ads using Stable Diffusion for image generation. For captions, we employ Retrieval-Augmented Generation (RAG) models, analyzing product reviews and Reddit posts to craft impactful, data-driven ad copy.

Our evaluation highlighted the effectiveness of our methodology, with promising precision, recall, and F1 scores from models like Pegasus and BART. 

Through this project, we gained hands-on experience with advanced AI tools and technologies. We learned to effectively use image generation models like Stable Diffusion to create professional-quality ads and leveraged LLMs for caption generation through Retrieval-Augmented Generation. Additionally, we developed expertise in setting up and using AWS infrastructure, including Pinecone for vectorization and storage, to enhance the scalability and efficiency of our pipeline. These insights not only deepened our technical understanding but also demonstrated the practical applications of these tools in solving real-world challenges in advertising.