A brief summary of my experience with one of Udacity’s Nanodegree programs

Motivation

Udacity offered a 1-month free promotion for many of their Nanodegree programs during March of 2020. In fact, it is now early May, and I believe that promotion is still going on, but I’m not sure for how long). Since I’m pursuing a career in Machine Learning Engineering, I thought I check out their MLE Nanodegree. I was able to complete the entire degree in the free trial period, and I wanted to jot down a few of my thoughts about the experience for anyone else who may be considering enrolling in a udacity course.

Content Summary

Software Engineering Best Practices

  • Basics of Object Oriented Programming
  • How version control can be used in data science
  • Test Driven Development
  • Creating your own python package that can be installed via pip

Using Amazon SageMaker

  • How to build models and deploy them using Amazon Sagemaker
  • Model Hyperparameter tuning with Amazon Sagemaker
  • How to make the model available for real-time predictions using AWS Lambda Functions and API Gateway

ML Case Studies

  • Population Segmentation
  • Plagiarism Detection
  • Building Custom Models with sklearn or pytorch
  • Sentiment Analysis

Capstone Project

  • Completing an ML project of your choice
    • includes a project proposal, report, and repository of the source code

Things I Liked:

Organization/Content

Overall, I thought the course was very well organized. It included plenty of relevant content that was covered in a logical order. Additionally, many of the sections included links to external sources for additional information if the student was interested in learning beyond the scope of what the course covered. The organization was a huge plus for me. In the world of programming, you can find content for anything and everything online for free, and so one of the primary benefits of a course is simply the fact that it curates much of the available information into a consumable format. Udacity did a great job at this.

Feedback

One of the things I really enjoyed about the course was the feedback that was quickly given for each of the projects. Within 24 hours of submitting a project I would get a response indicating whether or not everything was up to specs and if it wasn’t they would point out specific areas of concern along with actionable feedback.

Things I Didn’t Like:

Price Udacity normally charges about $400 dollars a month while you are enrolled in their degrees. While their courses are high-quality, I’d be hesitant to say that I think this is a fair price. With a little research and discipline you could find all of the material online for free. So you have to decide for yourself if the cost of gathering/organizing the course material is worth that much to you.

Side note: The fact that the course was temporarily free was a very effective motivation for me to finish the nanodegree quickly as I didn’t want to get hit with that high-cost at the end of the month. As such, I adjusted the suggested project deadlines so that they all fell within the 30-day free trial period and consistently dedicated 1-2 hours a day (occasionally longer) to the coursework so be able to complete it. Hand Holding

Another dislike I had was that I think a little too much of the dirty work for each of the projects was already completed for you. They provided a notebook or code base with everything mostly set up, and then I just had to add in components for the model training and/or tuning. As a budding MLE, I was hoping to get my hands dirty with more of the process. However, for beginners, it is nice to have working examples that can serve as useful references when building out your own solutions.

Restricted Access After graduating from the nanodegree, I was surprised to see that I lost access to all of the course material. While continued access was never promised by Udacity, I still think it’s a little lame that it isn’t included as one of the perks of completing the program. Fortunately, I took meticulous notes, and kept a spreadsheet with links to useful resources for each module so I didn’t come away empty handed. However, given that the course is really only a beginner course, It would be nice to have continued access to the material for reference as I continue on and build more complex solutions in industry.

Conclusion

As with most online courses, I think the promised benefit is far over-hyped in the marketing. This nanodegree was great for getting my feet wet with Machine Learning Engineering using AWS, but that’s about it. I wouldn’t consider myself an expert by any means, and If I was a recruiter I would be hesitant to hire anyone with nothing but a Udacity certificate as their relevant credentials. However, this course did have a moderate amount of rigor, and if you have the dedication to complete the nanodegree during a free trial promotion. I would definitely encourage you to check it out.