Azure Exploration & Certifications

I’ve been into Azure for quite a while now as it’s my cloud provider of choice. As can be seen on the following screen, in 2024, I loved <3 spending my monies on cloud services.

Azure costs

Initially, I did some experimenting with VMs on GCP, but this hyperscaler didn’t stick with me.
Iirc, I didn’t like the interface, and, frankly, Google radiates these woke vibes for a few years now. If you’ve been around, you know what I mean. The integration with Azure and the whole UI/UX is just way better if you’re coming from a VS, VS Code, or especially a C# .NET background.

I did the basic Azure certification AZ-900 for free back in early 2023, thanks to a Microsoft Virtual Training Day.
It was a good introduction to the whole ecosystem and built a solid understanding of concepts like IaaS/PaaS/SaaS, Serverless, Resource Groups, and Locks. After that, I stopped blindly provisioning Linux VMs and explored other concepts, which led first to WebApps, then to Containers, and later on to K8s/AKS. Yet, there were still prominent services I often saw in the portal that I never interacted with: Databricks, Custom Vision, Purview—what even is this stuff? Will I ever need this? I knew about Active Directory and Entra, but on most other non-developer-focused services, I had significant blind spots.

After years of using Azure in production environments, what could be better than diving deeper into it?
Six weeks ago, I decided to actively pursue Azure’s non-expiring base-level certifications.

I have a lot of industry-related certs — easily 30+ by now — and most of them are valuable, accredited ones.
They were properly proctored with exams and fees. Some even required me to apply or qualify before taking them.
A few I had to retake multiple times to pass. This means there’s a barrier to achieving them, which adds value.
Except for the PMP, though, these are all foundational-level certifications. They’re introductions — Pareto’s 80%.
E.g. I have a low-level Yellow Belt, but I haven’t done Six Sigma in years. Yet, being introduced to its toolkit and general terminology a decade ago gave me lasting insights into what it’s about.

What I love most about these foundational certs is that they don’t require recertification every few years, unlike the PMP. Recertification is costly—in both time and money —and frankly, it’s often not worth it. You never forget the main principles, though you might quickly lose the finer details if you’re not constantly applying them.

Three Azure Foundational Certs

To keep a long story short:

I completed the three foundational certifications in AI (AI-900), Data (DP-900), and Security (SC-900).
I passed the last one yesterday—a week ahead of schedule—and scored around 90% on each of them. Unfortunately, this time I had to pay for them since Microsoft discontinued its free exam program from the Virtual Training Days.

A Couple of Key Takeaways:

  • Custom Vision: An AI solution where you bring your own pictures and train an ML model to recognize your item(s) in them.
  • AI & ML Distinction: Azure distinguishes between classical AI/Machine Learning and new Generative Transformers (ChatGPT et.al.).
  • Audio Transcription: Use the classical Whisper model—a cheap neural net trained on hours of audio. Afaik, it’s used by YT and others for auto-subtitling.
  • Document Intelligence: An AI service that can OCR/parse documents for shared fields, extract them, and store them in a DB schema of your choice.
  • MS Intune: Manages endpoint devices like laptops and mobile phones. It integrates with Entra/Active Directory to remotely shut off or brick devices.
  • MS Purview: Helps manage and secure organizational data, preventing theft by insiders. Allows data labeling for retention/privacy and tracks USB copying.
  • MS Defender: There’s a gazillion different Defender applications covering everything from Storage to Endpoints.
  • Sentinel: An integrated SIEM/SOAR solution for securing resources.
  • Azure Synapse Analytics: Essentially a data warehouse but with a fancy name.
  • Big Data Solutions: Databricks and HDInsights are powered by Apache Spark under the hood.
  • CosmosDB: Not just one NoSQL database—it’s 5-6 types, from Graph to MongoDB and Columnar.
  • PostgreSQL: Expensive in Azure, no matter the setup. Meanwhile, all the Microsoft SQL DB variants are dirt cheap.
  • Data Integration: Import data visually with Azure DataFactory Pipelines or read on the fly with Polybase and dump it into Azure DataLake.
  • Blob Storage to DataLake: Simply enable Hierarchical Namespace to upgrade a dumb blob to a (dumb) lake.

I also skimmed through the curriculums for the Azure Developer Associate, Azure Administrator Associate, and the two expert-level certs and considered on taking them. Having worked on practical projects involving most of the topics in those certifications, I don’t expect to find much new. But in all fairness, I don’t think I’ll pursue them. It’s not the cost or effort, but as earlier mentioned the recertification requirement (every 12 months!) that turns me totally off. Otherwise, I might have scheduled them for 2025.

I still have the WTF Model to release this year (2024), but that’ll be another post. I plan to launch a minimal model for the WTF, which is already right now in testing as a static WebApp and seems to deploy successfully via my existing DevOps pipeline. However I’ll make an official announcement when it’s proper. In 2025, I also aim to sell access to the full version not just via Shopify (as I do now) but also on Steam.

comments powered by Disqus