What should be my roadmap after Machine Learning Specilization?

I have currently taken the Machine Learning Specialization course from Andrew Ng and have almost completed it. Afterward, I saw the Deep Learning Specialization course, but when I looked at its content, it seemed very similar to the first course I took. I’m not sure how much Machine Learning Specialization and Deep Learning Specialization courses resemble each other, and whether I need to take the Deep Learning Specialization course or skip some of the lessons. I want to create a roadmap where I don’t waste my time, but I’m stuck. My first goal is to become proficient in Tensorflow and machine learning, and then I’m not sure what goal to pursue next. Do I need to choose a field for artificial intelligence, and which DeepLearnşng.ai courses should I take? I would be grateful if you could help me with this.

Hi @Kerem_Boyuk,

You may find DLS C1 W2 familiar, but it is not a bad idea if you take that as a bridge to the rest of the specialization which is almost not overlapping with the MLS. The rest of the DLS C1 will focus on building a neural network from scratch but deeper (more realistic) than that in the MLS, and more importantly, we will also start discussing deep neural network techniques and their considerations behind.

As you move forward into C2 and C3, you will be going through many building blocks of a neural network and how people actually do them. Those are something you need to know to formulate your own recipe when you model a dataset on your own.

C4 and C5 of the DLS are for computer vision and natural language processing respectively, and you will have a chance to think through how some of the very popular and practical architectures worked given the lectures, assignments and what you have learnt from the previous courses.

DLS can give you a reasonable introduction to deeper and useful networks, which is some good foundation for you to explore an AI field in the future.

If you are not sure about the DLS, then audit the lecture videos first.


1 Like

I wonder when will I really be able to do projects and have this foresight? What should i do next, i need to complete roadmap, can you help me?

Hello @Kerem_Boyuk

What I can do here, for learner who posts a thread, is to lay out facts that are to my knowledge, or try to figure out some ways to let the learner think through their question themselves, although it is actually harder and might be less favourable than just giving out the answer which is sometimes my only best choice.

If you are looking for something more or something other than what I can do, including your personalized roadmap, you will need to find it in somewhere else. :wink: Before you find it, I don’t see any harm to start auditing the DLS, and then make your own choice of whether or not to complete it.


1 Like

Hi @Kerem_Boyuk

In my opinion the best personal sequence for you depends on:

  • where you stand now (beginner | medium | advanced)
  • what you want to achieve (e.g. become an expert in Tensorflow or computer vision and also build some knowledge in NLP)
  • your strength and background of the industry you want to work (e.g. background in image processing, working in automotive industry in the automated driving area with focus on sensor fusion and deep learning)
  • your timeline, considering how much time you want to invest in your learning roadmap (e.g. 5 hours per week for 8 months or so)

As some inspiration and ideas for your roadmap, feel free to check out these threads:

Hope that helps!

All the best for you and happy learning.

Best regards


Hi @Kerem_Boyuk,

Please don’t be discouraged by my previous message.

Looking for a roadmap is a right thing to do, but I am not the person who knows you well to really give you that. If you are not able to find that person yet, and if I were you, I would be my own guide. For example, I would make the most of what Christian has shared with us, and tried to come up with some ideas about a roadmap (not necessarily a full roadmap, but perhaps a short-term one) to make sure I am always on the move and not waiting. It might not be the best short-term roadmap, but it is not going to be very wrong if you have done your research seriously and taken into account your own interest.

Opportunity might come anytime but you will need to be able to grab it. That is why in the mean time, I really encourage you to make some plan for yourself and keep improving. Try more, do more, experience more, so that when you can speak with that right person, you can show that person your committment with a mature conversation backed by all of those efforts.

To me, self-learning is a trend, and many people live with it. New things come up everyday and sometimes school cannot catch up fast enough for students, so we need to learn them ourselves. This is the case no matter we have someone telling us what to do or not. To this end, we are all the same :wink: You might not have someone who can give you a roadmap yet, and I do not have someone who can give me a roadmap either.

Therefore, try to develop something (be it a plan or a short-term roadmap) for yourself so that you can expose yourself to new things, and you can practice practice and pracitce to strengthen your muscle memory on building machine learning models. That muscle memory is going to be very helpful no matter what AI field you will try in the future. The DLS is really a good compilation of recent deep learning development that can keep you busy, and it is free to audit the videos anyway :wink:


Keep learning, and cheers!


1 Like

Hey @rmwkwok

I understood exactly what you meant in your previous answer, but thank you again for trying to explain it. Indeed, I am alone on this path, at least thanks to deeplearning.ai, I can direct my questions to valuable mentors like you and get my answers. I will continue my roadmap with DLS as you said and then I will switch to MLOps. GNAs are the most interesting to me in my field because I think it is a very important and innovative step for AGI. I want to be one of the developers of this field. I want to make Machine Learning able to adapt to many data, not a single data. Thank you again for your reply.
Have nice day