Guide for Machine Learning, NLP, Computer Vision - (Roadmap) Feedback

AI Discussions
1

Hello everyone, Hello everyone,

I hope you’re all doing great! I’ve been working on an academic project for UX design, and as part of it, I had to create a new feature for any website. I thought DeepLearning.AI was a great place to apply my skills and knowledge.

I started by focusing on a common problem in the community:
“Where should I start? Is there any roadmap for learning AI?”

This is a question I’ve seen many people post here on the forum. So, I did some research and analyzed different opinions on the topic. Now, I’ve created a prototype in Figma (using the same design style as DeepLearning.AI) for a new feature where users can explore possible roadmaps for different AI fields, like Machine Learning, NLP, and Computer Vision.

Here’s the link to the prototype: https://www.figma.com/proto/patEqKNdBQIg3Sdq77y42e/DeepLearning-Redesign-Path?node-id=1-4&starting-point-node-id=1%3A4&t=BEDJxPl6oFLaKPST-1

Maybe you’ve seen my posts about this project before, but I haven’t received enough feedback yet. I can’t finish the project without your help!

Just to clarify: this post isn’t meant to promote or publish anything. My goal is to improve my UX skills and help others in the community who might need a guide for their AI career. I hope this makes sense.

Finally, I’d really appreciate it if you could give me some feedback and let me know:

  • What do you think about this idea?
  • Would you like to see a feature like this in the future?
  • How does the design look to you?
  • Do you feel the information is clear and accessible, or did you find anything confusing?

Thanks so much for your time and support! I hope this project helps many people and inspires you in your own career too.

1 Like
2

The overall idea of using a neural network like UI to show course dependencies is cool. I didn’t read the text part. Here are some points to consider for the diagrams:

  1. Why is deep learning specialization and MLOps part of machine learning specialization?
  2. When looking at the tab for NLP track, courses from tensorflow developer professional should be a pre-requisite. IMO, tensorflow developer specialization related courses should be included in the fundamentals for any specialization that needs it.
  3. Having one entry point per flowchart is less confusing. In computer vision specialization for example, have a single input to the 1st input layer or explicitly indicate that any starting point is valid.
  4. For computer vision, if advnanced tensorflow specialization (TF3) is part of the “specialization” section, shouldn’t the first 2 courses of TF3 be a part of NLP specialization as well? How about including them in the optional section?
3

Thank you so much for your detailed feedback! It really helps me improve the prototype and consider different perspectives.

Regarding the deep learning specialization and MLOps being part of the machine learning specialization, I included them there because they build on foundational ML concepts. However, I see your point, and maybe it would be better to clarify the relationships or reconsider their placement.

For the NLP track, you mentioned TensorFlow Developer courses as prerequisites. I partially agree, as they’re valuable for many specializations, but I’d like to keep the focus on core NLP concepts first and then suggest TensorFlow courses as optional or complementary, depending on the learner’s path. What do you think?

I like the idea of having a single entry point per flowchart—it definitely makes things cleaner. However, I also want to give learners flexibility, especially since not everyone starts from the same level. I’ll think about how to balance simplicity and adaptability in the design.

As for the advanced TensorFlow courses in the computer vision and NLP tracks, that’s a valid observation. My initial thought was to keep them focused within their respective specializations, but I see how including them as optional content in other tracks could provide more value. I’ll explore that further.

I really appreciate your input​:blush:

4

You’re welcome.

This is my understanding of the labels:

  1. Fundamentals: Basic requirements for completing the core topic of interest (eg. machine learning, computer vision etc.)
  2. Specialization: These courses are a must do to demonstrate competence in the core area.
  3. Optional: Additional courses to consider if you want explore further in the core area (eg. GANS for CV).

Tensorflow developer professional (TF1) courses can be considered as an optional fundamental requirement for NLP since DLS covers the basics (a different color circle / a separate block for optional fundamental courses?)

Adding @Deepti_Prasad to confirm if DLS sufficient or if TF1 required for completing NLP.

1 Like
5

Thank you @balaji.ambresh for adding me, actually I did come across this post, but didn’t read it through as I thought he was providing his experience to be an NLP expert.

Although I agree with @MarcoATL, people wanting to become NLP expert come with different level of experience, I would make prototype which mentions

a flow-chart, like if knows python programming, pytorch, basic tensorflow tf1 or general tf and keras, then probably @MarcoATL suggestions might fit in somehow.

But we actually never know how one makes a decision, so always provide the basic path to be an NLP expert mentioning significance of each course on how it might help in understanding NLP concept.

I have seen many experts here suggesting first MLS(Machine learning course) and then DLS course, but I kind of disagree on this because Prof.Ng has covered like every nuke and corner on how to handle data, first in DLS, so when I did MLS later, for me it was practice specialisation of what I learnt and understood in DLS.

For me TF1 would be necessary pre-requisite no matter you know python or keras or tf in general or not, to be an NLP expert. Here is my reasoning. TF1 was one of the best explained, very clear, easiest course which help me understand the concept i learnt about cnn, kernel, filter, data conversion, model training using tensor and keras from DLS which eventually will help in course 3 and course 4 of NLP specialisation to do more smoothly and understand on basic concepts as they updated with tensorflow version of assignment. NLP also convered sequence modelling which was easy go as it was thoroughly taught in DLS yet understanding of tensorflow helped me understand some of tensor data conversion flow and how inputs data is converted to tf type.

Now come to Computer vision which is part tf3 specialisation, this you could add as an optional course as I understand they aren’t related too literally but I personally love this course. it was the toughest course when I was naive in ML and AI, but practicing and teaching this, still makes the curious kid to explore more. My personal goal is healthcare, so for me computer vision was must as it cover Bounding box, transfer learning, object detection, auto-encoder visualisation which are must for any creator.

MLOps is again an advanced course which could be optional, to understand NLP.

Another thing @MarcoATL probably missed is adding the numerous selective short courses related NLP, as text, image and videos are converted into magical numerical for them to dance together, doing short course regarding LLMs is like I saw that concept in NLP, aa ha so temperature holds importance for the chatgpt to give me output relative to the deterministic or semantic, covering cosine similarity concept explained in NLP.

I have seen learners doing NLP even when they don’t know much about python, that’s a big no as they might end up upsetting themselves.

Learning becomes fun, exciting and growing when it is simple yet challenging, complex yet correlative for learner to understand and growing so they can create their idea from concepts of different specialisation they have learnt.

Namaste
DP

1 Like
6

What do you learn from here ?just how to use ai?

1 Like
7

Thank you so much for your detailed insights and suggestions! I’ve made sure to address your points, including emphasizing the importance of TensorFlow as a prerequisite for NLP and adding notes for learners with prior experience. The updated prototype is available through the original link. I appreciate your expertise!

8

Thank you for your thoughtful feedback! I’ve updated the prototype based on your suggestions. Specifically, I adjusted the diagram for better clarity and incorporated TensorFlow Developer specialization as part of the Fundamentals for NLP. Please check the updated version in the original link. Your input was incredibly helpful!

9

The notes look good.

It’d be nice if you could share both versions of the prototype to get a better understanding of what changed.

Please explain how you’ve changed the prototype to exclude DLS from MLS?
Since MLS mostly focuses on tabular data, it’s better to exclude DLS from MLS. Newbies are likely to get confused about why there are 2 specializations to complete when there’s already MLS in place.

The following courses from advanced tensorflow specialization are better in the optional section in CV:

  1. Custom models, layers …
  2. Custom and distributed training …

The above courses are also good to have in other specializations in the optional section where applicable.