The first item on my action plan is now formally accomplished and the first milestone on my new adventure is finally here. I took and passed the Designing and Implementing a Microsoft Azure AI Solution (AI-102) exam which certifies me as an Azure AI Engineer Associate! . Preparation for this exam was distributed in very small dozes as I’m currently working a full time job, plus the many other responsibilities a dad of 3 has. Basically, I studied a few hours at night and (some) weekends, airport waits and holidays.
These are the skills that, with a little bit of practice, I should be able to handle now:
- Plan and manage an Azure AI solution
- Implement content moderation solutions
- Implement computer vision solutions
- Implement natural language processing solutions
- Implement knowledge mining and document intelligence solutions
- Implement generative AI solutions
This is my shiny new Azure AI Engineer Associate certificate!
Here’s the link if you want to check it digitally:
Presenting the proctored exam online
I took the exam online using Pearson’s OnVUE service. It cost me $USD 140 since I’m based in Canada, US locations are currently $USD 160.
The overall experience was smooth, although there were a few minor hiccups to address.
Wired Internet requirement vs Showing your working area with the webcam
Pearson requires that you use wired internet to ensure your connection is stable during the test. Then, during the pre-check, they ask you to show them your working area with your webcam. Since I was using a laptop (webcam integrated) , I’d need to disconnect the cable to move the laptop around. Therefore losing connection with the proctor.
Who uses wired Web Cams?
I think this requirement will have to be rethought . I expect more and more people will be using laptops with integrated webcams now. There’s a possibility that, if your Ethernet cable is long enough , you could manage to move you laptop around to pane out your working area without risking disconnecting the CAT cable, but I don’t find that very feasible. Also, I’m not sure if losing connection with the proctor momentarily (while the laptop grabs the WiFi signal) would be a huge deal (i.e. get you disqualified) or if the system will re-connect when it detects a new signal and done deal. The instructions before the test alerts you that, if something changes like the Internet connection (IP?), you may get disqualified though.
The first proctor that got assigned to me wasn’t really open to understand my point and kept repeating ” I understand, but checking your working area is a requirement and I wont be able to release your exam if I don’t see your desk trough the video feed”. I kept explaining that, if I disconnected the Ethernet cable, well… I’d lose connection with her.
At some point, either she got frustrated or decided to escalate the case to another proctor who came up with the idea of asking me to use a small mirror to show him the desk without having to move the laptop. He told me I was allowed to leave my desk to go and get the mirror, so I did and showed him the area. He asked me to remove a few items (Rubix cube, 2 cups holding pens, a few figurines) I had on my desk and then released the exam.
All in all, it wasn’t a great disruption but an unnecessary one.
💡TIP: Use a wired camera or long ethernet cable if you’re following Pearson’s advise on using wired Internet connection
They don’t allow headsets but validate headset during pre-check
They don’t allow headsets, which makes sense and wasn’t really an issue. However, I don’t recall seeing this requirement in the rules. I removed my headset and used the integrated mic and speakers to communicate with the proctor. What I found a bit misleading was that, during the system pre-check, the system validated my Jabra headset as an acceptable audio device and mic. Perhaps they should update the system to remind users that headsets cannot be used during the exam and suggest switching to the integrated speakers and mic.
💡TIP: Headsets are not allowed. Even if Pearson’s system validates them during pre-check
It was way more granular than what I expected
I understand not all exams are equal and maybe they have several versions and randomly assign one to each student. Mine, went very granular in regards to the name of the functions and code in SDKs and API calls. I wasn’t expecting that but I seemed to do well in understanding the logic behind the questions.
💡TIP: Do a quick review of all the code involved in the Labs so you have fresh in your mind the most common model and function names.
My study process – Azure AI Engineer Associate
- I went trough the Self paced Microsoft course
I find this material very comprehensible and thorough. - Used Google Slides to take notes
Sometimes, with some topics, I need to visualize the information I’m getting. Scramble it, move it around and use my own logic to really get the point.
NOTES
Here are my notes in case you want to take a look.
👇
- Took Microsoft Practice Assessment a few times.
- Watched the Exam AI-102 prep videos.
Note: I didn’t find it very useful. Not much more elaboration than reading the titles of the topics on the curriculum. - Watched John Savil’s amazing study cram
https://www.youtube.com/watch?v=I7fdWafTcPY&list=PLH8AhFMOOrcR6xnWHoMuapVOX6AuQzi8U
What’s Next?
As shown in my action plan, my next step after getting my Azure AI Engineer Associate certification is to develop a project powered by Azure AI Services and add it to my portfolio.
I’ve been brainstorming with ideas. My perfect first project should meet the following criteria:
- Low cost.
- Time efficient.
- No doorkeepers.
I don’t have the time to prospect the project to a company and go through their decision making and approval processes. That would take a lot of time, multiple meetings with stakeholders, and eventually it will be very sensitive to budgets. I’m planning to keep this one private, pay for my development expenses (i.e., paid AI services, hosting, etc), document everything and then kill all the paid services after I finish. - Trained with publicly available data.
to make the first bullet point possible I need to use easy to access, publicly available data
Stay tuned!
AI Hero