Manifest AI

The Manifest-AI project for Modern Mantra was an ambitious exploration into the realm of generative AI, particularly focused on creating high-resolution “collages” through a unique, iterative approach. In partnership with an AI engineer, we capitalized on the robust capabilities of GPU-equipped EC2 instances, setting the foundation for a pioneering prompt-based generative image pipeline.

Central to achieving a truly decoupled service architecture for this project were several AWS services that worked in concert to manage and process long-running jobs, which could take 20-30 minutes for a single asset. Amazon SQS (Simple Queue Service) was integral to managing communication between distributed components of our architecture, ensuring that messages regarding job status and completion were reliably transmitted. Amazon SES (Simple Email Service) facilitated notifications and updates to stakeholders, providing alerts on job progress and completion.

Amazon S3 (Simple Storage Service) served as the backbone for storing input data and the generated high-resolution images, enabling scalable and secure access to assets throughout the processing pipeline. The deployment of these services, alongside our primary EC2 instances, exemplified a sophisticated approach to handling intensive computational tasks. This architecture allowed for scalability, reliability, and efficiency, ensuring that even the most resource-intensive jobs were completed smoothly without overburdening any single component of the system.

Although the project concluded in the pilot phase due to external factors, it underscored the potential of combining cloud computing with AI to push the boundaries of creative image generation. This initiative was a testament to the power of leveraging advanced AWS services in concert to solve complex problems, providing a scalable and flexible infrastructure capable of supporting innovative AI applications.