Access to AI

Overview

Wait... doesn't the cloud solve that? Sort-of. We have seen various organizations, ranging from startups to mid size companies and even large computing centers struggle with provisioning AI resources both in-house and in the cloud. Cost is a big issue: as demand has grown exponentially, so has the cost of high-end high accelerator cards. Then there's access: in the cloud, in times of high demand, there are just no instances to be had; in-house operations can find that ordering cards may involve a long lead-time, provisioning them is difficult, as is estimating demand. Then there's data: training AI models requires tons of data. If you write out, or worse save the data your generating, you're dead.

Our Strategy

AreandDee's approach seeks to lower the infrastructural entry cost and avoid expensive bottlenecks.

• Work with large, established resource providers to obtain block allocations of resources.
• Work with startup resource providers like MyTorch.Net, to obtain interactive remote access to powerful AI systems.
• Minimize data production: focus on continuous training approaches that train models as the model runs.
• Focus on tools, like Heal-Pics, that can visualize large geospatial datasets in seconds.

Current Progress

Information about our infrastructure journey will be displayed here.