Has anyone certified in Amazon’s Machine Learning path? If so has it been helpful implementing your Fast.ai projects into the Amazon environment?
Not Amazon but I have done the Google Data Engineer certification. They focus on Googles ML, DB, streaming and other services and Tensorflow. It helps if you want to deploy a service into Production as there is more to it than just training models.
1 Like
It depends on what services you want to get familiar with from Amazon. Their ML Learning path seems to have a good amount of background on ML, which you may not need if your goal is to learn more about how to utilize their environment.
Looking at materials to prepare for the developer or cloud architect associate exams, which get you familiar with the “basic” Amazon services (EC2 and EBS, S3, Lambda, IAM roles). The one thing that the ML path would focus on more than the others would be SageMaker, but I’m sure there is a video series just for that if you are interested.
(As for my background, I have the AWS Developer Associate and Bid Data Specialty certs. I’m planning on taking the ML Specialty exam sometime in the next few months)
1 Like
Great! I have AWS Architecture associate and was planning on taking their Security and Big Data. When I saw ML, I think I will swap that out for big data in my plans. Was planning on using Cloudguru again for their training videos.
Let me know how it goes or if you have some good training resources!
Agree, I think the ML exam would make more sense. The Big Data was quite difficult - I was surprised I passed. I used ACloudGuru and LinuxAcademy to study for the Big Data Exam and found LinuxAcademy to be better (ACG seemed to be outdated for the Big Data Specialty).
If you already have a good grasp of ML concepts, I would think that exam is going to be much easier than Big Data. I haven’t yet reviewed the courses for ML, but will try to remember to share thoughts after I take the exam.
1 Like
Completed my AWS security and moved onto the AWS Machine Learning cert.
Did acloudguru and it did help me think of ways to improve putting my fast.ai learning into production. The example and labs are pretty good. However, I am still not getting a good sense of what the questions would be, and some areas seem goofy.
For example, they discuss feature engineering to get better results with a case of CNN’s of turning the mnist series into a grid of 1s and 0s. I would consider that less feature engineering and more about how CNNs operate. They should discuss image transformations to improve results.
Also, I am having trouble with Sagemaker. Their example is excellent for setting something up in production. The acloudguru examples with UFOs are a nice touch. But when I am trying to use sagemaker for a Kaggle Fraud dataset, I am running into errors which are hard to problem shoot and cost money for doing so.
Overall, I think it is excellent as I have a model I am trying to get into production at work, however, since I am having problems on a smaller dataset, not sure how I can get it to scale while staying in budget.
Also, need to find some better practice tests for this exam.