Mapping the AI Stack
As many of you know, over the last couple of years I’ve been learning about AI in bits and pieces. Most conversations I’ve come across focus on just one slice of the landscape. But being a process person, I wanted to see the whole system end to end. So I mapped it out for myself.
At a high level, I see three broad layers:
Infrastructure: Hardware, Compute, Data, Development
Models: Foundation, Specialised
Applications: Vertical, Horizontal
As I zoomed in and out of this stack, what stood out to me is how choices in one layer ripple through the others, almost like a bullwhip effect from operations. A resource constraint in infrastructure doesn’t just stay there; it can shape model training, evaluation, and eventually what applications deliver to the end customer (enterprise or consumer).
In the schematic above, I’ve mapped a few names of companies that you may have heard of, just to make sense of what they do. This map helped me connect the dots. Sharing here in case it helps others who are also trying to step back and see the bigger picture.
As I dig deeper, I’ll keep sharing my learnings here. Feel free to subscribe, unsubscribe, mute or close the tab. Whatever keeps your nerves steady :)