Reducing Cloud costs for an AI Medical Tech company by over 95%

Problem

A US-based MedTech company approached us with the requirement to build a scalable cloud-based solution for running Machine-Learning experiments by partner laboratories throughout the US East Coast, US West Coast, and Western Europe. Existing solutions weren't scaling well, and each lab had its own internal software solution, which cost significant time & resources to maintain. In addition, due to the fragmented nature of their solutions, there was a substantial delay for processes like data collection and global system analytics.

Solution

Early through our investigation process, we noticed a major savings potential due to limited working times. The staff at the labs was only conducting experiments at certain times of day, and only during weekdays. This meant that the infrastructure was only used ~8.3% of the time.
We proposed and later implemented a stack of scalable server fleets, with a specialized trigger to automatically turn them on, and potentially scale them when needed.
As suspended servers only incur storage fees, which are very minimal, we were able to substantially reduce the cost of the entire architecture. In order to further ease infrastructure management for the researchers, we've developed a simple UI that would allow for experiement data to be uploaded, while facilitating optimal storage and compute usage.
Total cost savings of over 95.15%