August 2024
Professional Services Firm (Big Four)
GenuineBee collaborated with a leading professional services firm, part of the “Big Four” in the US, to revolutionize their employee profile photo management process. The project aimed to automate and centralize the management of these photos, primarily focusing on background removal and replacement for uniformity. GenuineBee devised an innovative solution leveraging AI technologies, successfully addressing the client’s unique challenge.
One of the "Big Four", a leading professional services networks in the world, this company has offices in more than 150 countries, more than 270,000 people and annual gross revenues of more than US$40 billion. With an extensive online presence, their main global website alone contains more than 50,000 pages.
The client's US arm, comprised of approximately 20,000 individuals, was facing a challenge with their employee profile photo management. Multiple versions of the same photo and inconsistencies in photo editing resulted in repetitive work for the design teams and a lack of centralized photo management. What they wanted was to maintain uniformity in the photos by replacing the original backgrounds with a consistent, branded backdrop.
With the implementation of this system, we not only made the work of the design team more efficient but also added a touch of high tech to the employee profile picture management process. After all, who knew that keeping employee photos consistent could turn into a sophisticated operation involving cutting-edge AI and parallel processing? It's a reminder that even in the most mundane tasks, there's room for exciting innovation.
GenuineBee partnered with the client to analyze the problem thoroughly and design a custom end-to-end solution, one that would use agile methodology and incorporate the client’s brand design systems. We proposed using advanced AI techniques, specifically a method known as "image matting" for precise background calculation and extraction, accompanied by a message queue architecture for scalability. To ensure we achieve the best possible solution, we have employed three different AI models in parallel — PP-MattingV2-512, PP-Matting-512 and MODNet-ResNet50_vd — and we are constantly evaluating their performance to see which generates the best results. The client will then use that winning model going forward.
The solution boasts a robust AI-powered background removal feature, a scalable architecture, and seamless integration with the client's existing systems. During the testing phase, the solution received very positive feedback, and it is expected to save hundreds of design team hours once fully implemented. This means that the design team can now focus on delivering quality creative work, rather than repetitive editing tasks.
The technology stack for this project included Angular, Node, MongoDB, and Asus Storage. AI models PP-MattingV2-512, PP-Matting-512, and MODNet-ResNet50_vd were used for the image matting process. RabbitMQ was employed for the messaging queue architecture to ensure scalability and reliability. This solution was also integrated with other systems developed by GenuineBee for the client, including a personnel search service.
