Let’s get one thing straight before we dive into the technical depths of cloud infrastructure. The era of the “demo AI” is officially over. It was entirely acceptable a couple of years ago when you could wrap a basic open-source model pipeline in a lightweight web framework, slap it on a single graphics processing unit, … Read more
I’ll never forget the Monday morning I logged into an AWS billing dashboard to find a $12,000 surprise waiting for me. A client’s data science team, eager to test a fine-tuned 70-billion parameter model, had spun up five massive hardware-accelerated endpoints on a Friday afternoon. They ran maybe ten inferences, patted themselves on the back, … Read more
The tech industry loves to quote that famous Amazon study—the one claiming every 100-millisecond delay costs them 1% in sales. Frankly, it is an exhausting and ancient metric from an era when we were just rendering basic HTML over 3G networks. Let’s talk about the reality of building systems today. If you are building globally … Read more
Let’s get one thing straight right out of the gate. You cannot take your highly optimized, beautifully orchestrated AWS or Azure architecture, point it at a Beijing region, and expect it to work. I’ve spent the better part of a decade leading enterprise cloud migrations into mainland China. I’ve sat in enough late-night war rooms … Read more
Look at your cloud bill right now. If you are running a data-intensive architecture, I can almost guarantee two things. First, your object storage costs are significantly higher than your original projections. Second, your data transfer (egress) fees are making your finance department sweat. Object storage is no longer just a static bit-bucket where you … Read more
I’ve spent the better part of the last decade architecting, rescuing, and occasionally performing emergency surgery on cloud infrastructure for Fortune 500s and hyper-growth startups. If you’re reading this, your engineering team is probably staring down the barrel of a costly infrastructure inflection point. You already know the cloud computing landscape has moved far beyond … Read more