Not “5 tips for AWS.” Long-form essays on real production problems we’ve hit, the trade-offs we made, and what we’d do differently. Written by M. Talha, Maryam, Hassan, and David — never ghost-written, always specific.
A field essay on why Bedrock cost compounds non-linearly, why a 9-month migration window matters more than the eventual savings percentage, and what we got wrong on our first vLLM cutover.
The first call should reduce uncertainty for both sides, even if the answer is that you should not hire us.
A proposal without the person doing the work is not a plan. It is inventory management dressed up as strategy.
Good AWS account structure is boring, explicit, and understandable to both engineers and auditors.
Pager noise rarely starts in PagerDuty. It starts when nobody owns the difference between symptoms and user impact.
Savings Plans help only after the workload is shaped correctly. Buy commitments too early and you lock in the waste.
You do not need GPT-4 for every request. Most production AI bills are paying for the difference.
Active-passive sounds simple until you actually run a failover under production load. Five things the docs gloss over.
On-call is not just about page count. Three other knobs matter more for whether your rotation is sustainable.
Most companies measure cost per user. The number that actually matters is the slope of the cost curve relative to the slope of the user curve.
Same posts, organized by practice area. Useful if you’re searching for something specific or want to see the body of work under one of our four practice areas.
One email a month with the latest essay, plus production case notes we don’t publish on the blog. The only follow-up email we send. No marketing sequences, no “just checking in,” no upsell on services. Unsubscribe with one click.