I’m only weeks out from recertifying my AWS Solutions Architect Professional for the second time, and most of the studying is sitting a three-hour practice exam after work and then spending another few hours going back through it afterwards. It’s an architecture exam so it does test patterns, but they’re AWS-specific service combinations, the deep single-vendor recall of one stack, and my actual days are spread across AWS, GCP, Azure, on-prem and a pile of AI tooling, jumping between Windows, Mac and Linux. The exam wants me deep in one place, and real work has been taking me wide for years.
The first time I did this cert most of it was new, so the hours felt like they were getting me somewhere. This time almost none of it is, most of it has held, and what I’m actually paying for is re-proving it rather than learning it. There’s no shortcut, the exam is as long as it is, and fitting it around work has stopped feeling like studying and started feeling like maintenance on something I already own.
The direction I think matters next is running models locally, on my own hardware, and I still haven’t started, because the hours that would go there are going into re-proving AWS instead. That’s the stuff I actually want to be getting deeper on, the hands-on, figure-out-why-it-breaks kind of depth.
Real work keeps pulling across platforms while the recertification asks for depth in one.
I never knew what to specialise in and spent years as a jack-of-all-trades, which got me a long way but never got me the deeper knowledge that being genuinely expert in one area does, and I watched people who’d gone narrow get further on the back of it. Moving into solutions probably looks like the cop-out version of that, getting away with knowing a bit of everything instead of committing to one thing. But the breadth turns into its own kind of depth, because the thing I ended up deep on is how the pieces fit across platforms and what breaks at the joins when an AWS service has to talk to something running on-prem.
The recert can’t really get at that part. It can test whether I know an AWS service exists and whether I’d pick the right one for a scenario, and that’s most of the exam. What it can’t test is whether I know how that service behaves once it’s wired into an Azure identity layer and an on-prem database and whatever else the real estate is running, which is the work I actually do. So the hours keep going into re-proving one vendor, and I still haven’t started on the local models.