Insights Cloud & Dashboards
Cloud infrastructure and dashboards that actually get used
Cloud infrastructure fails businesses in two directions. Either it does not scale when you grow, or it scales and the monthly bill becomes a recurring surprise. And the dashboards meant to give you visibility? Half of them get opened once during the demo and never again. This guide is about avoiding both traps: choosing the right cloud, keeping costs sane, and building dashboards people actually reach for every morning.
In this guide
AWS vs GCP vs Azure: how to actually choose
People agonize over this more than they need to. All three are excellent, and for most businesses the deciding factor is not raw features, it is what your team and tools already use. My rough rule of thumb:
- Azure, a natural fit if you are already a Microsoft shop (Office 365, Business Central, Active Directory).
- GCP, strong for data engineering, analytics, and machine learning workloads.
- AWS, the broadest catalog of services and the largest talent pool, a safe default for almost anything.
The best choice is the one your team can operate confidently. A slightly less "optimal" platform that your people understand beats a theoretically perfect one that nobody can maintain.
How to avoid surprise cloud bills
Surprise bills almost always come from the same handful of causes, and every one of them is preventable:
- Over-provisioned, always-on resources. Big servers running 24/7 for a workload that spikes for two hours a day.
- Forgotten environments. A test setup someone spun up and never turned off.
- No budgets or alerts. The bill is only discovered at the end of the month.
The fixes are straightforward: right-size to your actual stage, lean on serverless for spiky workloads, and set budgets with billing alerts so cost is something you watch, not something that ambushes you. Good architecture is as much about cost as it is about performance.
Cloud architecture should be designed for your actual stage and budget, serverless for a lean startup, containerized infrastructure for something more established, not for a diagram that looks impressive.
Serverless, in plain terms
Serverless means you run code without managing servers. You hand the provider a function; it runs that function when something triggers it and charges only for the seconds it actually runs. Nothing sits idle burning money. For lean teams and event-driven or variable workloads, it is often the most cost-effective way to build, which is exactly why I used a serverless architecture on AWS for PSL Group's LLM research assistant: it scales up when medical teams query it and costs almost nothing when they do not.
Dashboards people actually use
A dashboard is only valuable if someone opens it on purpose. The ones that get used share a few traits, and the ones that die on the vine share the opposite:
- Built for a specific person. A dashboard designed for the portfolio manager who checks it daily beats a generic "executive overview" that tries to please everyone.
- Answers real questions. It shows the handful of numbers that drive decisions, not forty charts nobody asked for.
- Fast and current. Slow, stale dashboards get abandoned. Fed by a reliable data pipeline, the numbers are trustworthy and up to date.
Tools like Streamlit (for internal, data-heavy views) or React (for polished product surfaces) both work, the tool matters far less than whether it was built for the people actually looking at it. At OMERS I built Streamlit dashboards that portfolio managers used daily, precisely because they were designed around their real workflow, not a one-time presentation.
Quick cloud health check
Tick anything that sounds like your current setup.
Wondering if your setup is costing more than it should?
Tell me what you are running and what you are trying to see. I will give you an honest read on your cloud architecture and whether your dashboards are earning their keep.
Ask me directlyFrequently asked questions
Should I use AWS, GCP, or Azure?
All three are excellent. The right choice usually depends on your existing tools and team skills rather than raw features. Azure fits naturally with Microsoft-centric organizations, GCP is strong for data and machine learning, and AWS has the broadest service catalog and talent pool. The best pick is the one your team can operate confidently.
How do I avoid surprise cloud bills?
Right-size resources to your actual stage, prefer serverless for spiky or low-volume workloads so you only pay when code runs, set budgets and billing alerts, and review usage regularly. Most surprise bills come from over-provisioned always-on resources and forgotten test environments, both of which are avoidable with good architecture and monitoring.
What is serverless, in simple terms?
Serverless means you run code without managing servers. The cloud provider automatically runs your function when it is triggered and charges only for the time it runs. It is ideal for lean startups and event-driven or variable workloads because there is nothing to keep running, and pay, for when nothing is happening.
What makes a dashboard actually get used?
A dashboard gets used when it is built for the specific people who look at it daily, answers the questions they actually have, loads fast, and shows current, trustworthy data. Dashboards fail when they are built to impress in a demo rather than to be a tool someone reaches for every morning.