Tuesday, November 21, 2023
HomeAIWhat is a “cloud AI engineer”?

What is a “cloud AI engineer”?


  1. What cloud AI has made possible
  2. Is there a “qualification” to measure the degree of mastery of cloud AI? !
  3. What is Azure 100
  4. Failed the first exam. Reflection and countermeasures for the second test
  5. What is before and after qualification
  6. review the exam
  7. at the end

1. Cloud AI made possible

The word “cloud” has become widely known, but did you know that AI services and machine learning also have cloud services?

In the past, there were many hurdles to in-house AI model development, such as specialized knowledge, large amounts of learning data, and building the necessary development environment. Today, however, cloud services offer a wide range of functions for typical tasks in AI model development. By making good use of these, we are now in an era where we can develop in-house AI services that could only be built in a place with specialized knowledge and a well-developed development environment.

Typical services that utilize cloud AI services include the following.

image task Typical service name Case Study
image task
  • Computer Vision (image/video analysis)
  • Face (face recognition)
  • Immediate scoring system using Computer Vision for photo contests@impress
  • Photo tagging support system using Face @ FUJIFILM Software
  • Translator
  • QnA Maker (Q&A)
  • Automatic adding English subtitles for e-Learning teaching materials using Translator @Hokkaido University
  • Chatbot for healthcare professionals using QnA Maker@Chugai Pharmaceutical
  • Speech to Text
  • Text to Speech
  • Voice conference support system using Speech to Text@NTT DATA
  • Voice dialogue system with characters using Text to Speech @Gatebox
decision making
  • Anomaly Detector
  • Content Modelator (offensive content detection)
  • Anomaly detection system using Anomaly Detector

(Source: Azure Cognitive Services official page; Skill up AI research analysis)

In this way, services that can build AI models that can be used in various fields are provided. In other words, using cloud services for AI development will lead to a significant reduction in development man-hours, making it possible to test and operate AI services at low cost.

2. Is there a “qualification” to measure the degree of mastery of cloud AI? !

In this way, while the hurdles for in-house AI production have been greatly lowered, the usefulness of the skill of “selecting the appropriate cloud service for the AI ​​model you want to create” among various options is increasing. . There are famous cloud services such as AWS provided by Amazon Web Services, Microsoft Azure provided by Microsoft, and Google Cloud Platform provided by Google.

Each company prepares qualifications, but this time, I was interested in Azure, which is rapidly expanding its market share, so I took the Azure certification exam called AI-100. Report how your skills have changed. My specs at the time of testing are as follows.

  • I have a JDLA E qualification and understand what AI can do
  • Science graduate student conducting applied research on AI
  • Cloud beginner (I have used Azure DB service, but I don’t know design or operation)

3.What is Azure 100?

Here is an overview of the AI-100 exam (I can’t write about the details of the exam because a non-disclosure agreement is signed at the time of taking the exam, but I’ll give you a rough introduction).
Also, here you can see what is expected of candidates as an outline of exam skills, which translates to:

Candidates for the exam have expertise in building and implementing Microsoft AI solutions, including natural language processing, speech, computer vision, and conversational AI, using cognitive services, machine learning, and knowledge mining. is needed. An Azure AI Engineer’s responsibility is to analyze AI solution requirements, recommend appropriate tools and technologies, and design and implement AI solutions that meet scalability and performance requirements. Azure AI engineers translate visions from solution architects and work with data scientists, data engineers, IoT specialists and software developers to build complete end-to-end solutions. Candidates for this exam should have knowledge and experience designing and implementing AI apps and agents using Microsoft Azure Cognitive Services, Azure Bot Service, Azure Cognitive Search, and Azure data storage. Additionally, candidates should recommend solutions using open source technologies, understand the components that make up the Azure AI portfolio and the data storage options available, and when to develop custom APIs to meet specific requirements. must understand.

The above is an abstract description, but the breakdown of specific question fields is shown in the table below.

Analysis of solution requirements 25-30%
AI solution design 40-45%
Implementing and monitoring AI solutions 25-30%

The examination time and examination fee are as follows.

Test time: 210 minutes
Exam fee: 21,103 yen Languages
: English, Japanese, Simplified Chinese, Korean
Format: Questions with multiple choices and multiple answers
(*As of November 4, 2020)

4. Failed the first test. Reflection and countermeasures for the second test

Actually, I couldn’t pass the exam on the first try, so I took the exam twice. I passed with 700 points, but I failed with 680 points.

If you fail the exam, you will not be notified of the detailed score distribution. However, the three skill areas that performed poorly on the detailed exam are returned as a report. I tried to take measures based on this, but since only the “area” is shown, there is a problem that I have to identify what is included in that area. It was a point that I had a lot of trouble with.

For example, the report says that the following “Design solutions that include one or more pipelines” are poor grades. Since this is the only information I have, I have no choice but to steadily fill in the gaps in my knowledge while searching Google for “define an AI application workflow process” + “Azure”.

In this way, we search for items one by one. It took a lot of patience because I had to identify what I was lacking and go learn on my own.

In addition, the “learning path” is a collection of units and modules for understanding the outline and usage of Azure’s cloud AI service, but the “learning path” is an overview for the purpose of the summary It’s just an overview of, and there are quite a few. For example, if the service “Azure Cognitive Services” is stipulated as a “learning path”, in order to pass the exam, it is necessary to read the entire document and have enough knowledge to actually use and operate it. Therefore, it is necessary to have a point such as “getting a hit” instead of touching all the learning paths (I took it once, so I got a hit from the questions that were asked). Then I managed to take the exam for the second time and passed with a score of 780!

5. Before and after qualification

After taking the exam, I feel that I am able to:

  1. I was able to learn the concept of the cloud and know the main services and features of Azure.
  2. I have the ability to decide which AI service to propose for which issue
  3. Through hands-on, cloud AI can now be used for language translation, speech recognition, etc.

1 and 2 may seem mediocre, but “knowing” often leads to ideas for solutions to problems that you want to solve, and “knowing” and “not knowing” are very different. I thought.

In addition, 3 refers to the fact that by learning ” translate text and speech ” in the learning path, I was able to use Azure to run cloud AI that performs language translation and speech recognition. increase. In addition to learning knowledge on the desk, the AI-100 exam allows you to output knowledge while inputting it, so I think it will be a good experience to flexibly play with language translation that you can touch. .

I am a graduate student, but if you are a business person, by obtaining this qualification, you will be able to “think about the optimal AI solution in the business field” and “get in touch with engineers” regarding cloud AI. I felt that I would be able to do things like “I can answer questions from customers myself.”

6. Looking back on the exam

As I felt when I took the E qualification exam, there are limits to self-study, and I felt that it was not the best method for me. The Azure certification exam does not have a dedicated eligibility book, so I got through it by retaking the exam this time, but it takes a lot of energy to collect and read information on the Internet, so for those who are going to take the exam. There are a number of places where it is likely to get stuck in a swamp.

  • I’m interested in cloud AI, but I don’t know what to study
  • I want to learn efficiently in a limited amount of time.
  • When I hit a wall, I want to consult someone with knowledge
  • I want to pass the qualification together with people around me

For those who say, I think that taking advantage of specialized courses ( AI-102 courses, etc.) is a shortcut.

In Skill Up AI’s ” Azure AI-102 Compatible Cloud AI Solution Practical Course “, you will learn basic knowledge of the cloud and various Azure services through preparation videos, and the face-to-face course will focus on hands-on lessons based on the preparation content. . On the day of the event, the focus is on hands-on, making it a practical course where you can actually feel how to build an AI solution for actual problems rather than simply taking qualification measures. In addition, since the instructor is a certified trainer from Microsoft and an AI instructor for skill improvement, it is attractive that you can learn carefully through hands-on.

Unfortunately, this course did not exist at the time I took the exam, so I was not able to take it. It was also a difficult point that I had no choice but to solve it myself, so if there was an environment where I could freely ask questions to professionals who were familiar with both the cloud and AI, I think I would have been able to understand them efficiently.

7. Closing

Cloud AI realizes the implementation of AI without the expertise of engineers in the field of machine learning, which was a hurdle to the introduction of AI, and is making remarkable progress as a service that anyone can easily use. Now that AI has become more accessible through cloud services, whether or not we can utilize cloud AI at this point in time will determine the difference between career and business opportunities for individuals.
Although the technology and qualifications are not well known yet, I would like to use my cloud knowledge as a weapon in my future career!


Leave a reply

Please enter your comment!
Please enter your name here

Recent Posts

Popular Posts

Most Popular

Recent Comments