Choosing Your Path: AI Professional Program Course Selection Guide

Key Takeaways

Hello, my name is Armando. I am a course developer and course facilitator for the AI professional program. I'm a Stanford grad and I've been working with the AI professional program for over six years developing many of the course offerings that you'll be seeing in today's video. Um, in today's video, we'll be going over how to create a focus area within the AI professional program. Professional courses differ from graduate courses as they are two separate programs operating under the same umbrella. However, you can still take one graduate course and count it towards a professional certificate. Otherwise, we require three courses to be completed to earn our professional AI certificate. Our courses are adapted from the original onampus grad courses, maintaining the same level of rigor while being adapted for a professional audience. Enrolling in the program consists of a short application mainly to confirm your proficiency in some of the prereks such as calculus, linear algebra, and probability. All courses are fully online and consists of a variety of coursework ranging from coding, written assignments, etc. Each cohort is 10 weeks long and roughly estimates around 10 to 15 hours uh per week. Even though this can change based on your individual strengths and backgrounds, uh learners will have a personalized remote support from course facilitators via Slack and they have the option to engage with peers on there as well. Uh courses are pass no pass. Um and so at the end of the course, you'll have to reach a minimum number of points uh required uh in order to receive a digital certificate for course completion. Once you've successfully completed three courses, you'll receive the professional certificate. We have eight different offerings and are continuously exploring new um additions based on demand. Uh the content is derived from the graduate course content taught by Stanford faculty. So the rigor remains the same. We regularly talk with faculty to update after each iteration. Um and our core center around various different branches of AI. So it can be taken in any particular order of of of your interest. Um and they range from theoretical to practical. Um, and they're also grouped by different specializations. In a moment, we'll dive a little bit deeper into what all this uh entails. In the next part, we'll be going over the course offerings that we have within our program, starting with XCS221. XCS21 221 is a foundational AI concepts and algorithms course. Um, and we try to teach it in a through gamified assignments. So, you'll be building like Pac-Man, chat bots, core scheduling, and self-driving car algorithms, just to name a few. um you go over the history of AI and ethics and you just learn about the basic uh building blocks and terminology you'll need to be successful in other courses. Um this course assumes that you have some basic linear algebra and probability under your belt. Um assignments will primarily use numpy and does not really dive into deep learning. So you won't be using pietorrch uh in any way. Uh there's a total of seven assignments. Six are required as and one is optional and it's an ideal course for beginners who are looking to refresh foundational knowledge in artificial intelligence. Next, we have another foundational course XCS229 machine learning. This is a great course for those who have a strong math background and are seeking an introduction to ML. Uh the course focuses on building core ML algorithms from scratch using numpy. You'll be building algorithms like gosh and kernels, PCA, SVMs and so forth. Um the course is very proof and theor the theoryheavy um with an emphasis on statistical and mathematical comprehension to help you just better understand more deeply some of these um ML algorithms. Uh the prerex here is having a strong foundation in linear algebra probability calculus and some basic familiarity with some of the ML concepts um and experience with both Python and NumPy. Uh there's a total of five assignments for this course. Next, we'll be going over XCS224N. Um, XCS224N is a course that is great for those who want to get um more aware of the state-of-the-art models for natural language processing. Um, you'll be covering important models like transformers, llns, word vectors, and machine translation models. Um, it's a mix of theoretical and practical. um you'll be using a lot of PyTorch uh to help implement some of the deep learning frameworks for the NLP algorithms discussed in the class. Um the prerex here is to have some strong knowledge and probability linear algebra and calculus with familiarity with deep learning frameworks like PyTorch and there's a total of five assignments for this course. Next we have XCS234 reinforcement learning. Um this is a course that is rigorous for learners who want to have a really strong understanding in RL while also still exploring its applications and techniques. Um it covers real world RL applications such as RLHF which is a big topic in LLM tuning today direct preference optimization and model free policy evaluation. Um there is a heavy emphasis on both the math and statistical foundations for the above. So it's a proof heavy and a coding heavy course. Um so we for the prereex for this course um proficiency in probability stats linear algebra and calculus are definitely a must and having knowledge of optimization networks um especially deep learning frameworks like pytorches is going to be crucial for success and there are a total of five assignments. Next we'll be going over the XCS236 deep generative models course. Um, XCS 236 is uh those who have some experience with ML and are looking to deepen their understanding of generative models and its general applications. Um, generative modeling techniques uh will be explored such as auto reggressive models, GANs, autoenccoders and diffusion models just to name a few. Um, it's a mix of theoretical and practical uh but using very manageable data sets to to be able to understand some of the deeper algorithms. Um having a solid foundation of ML and uh deep learning frameworks is important. Um and familiarity with PyTorch is encouraged and there are a total of four assignments for this course. Next we have XCS224W machine learning with graphs. This is a course that is moderately rigorous for individuals looking to tackle large scale graph analysis focusing on data mining strategies to extract insights from complex networks. Um there's a hands-on practical feel to this course. Uh you'll be working on collab notebooks and trying to experiment with some of the algorithms discussed. Um and it very much highlights industry level applications. Familiarity with linear algebra probability and basic machine learning concepts is encouraged and experience with the with PyTorch is a plus and there's a total of five Google collabs for this course. Next, we'll be going over XCS231N, deep learning from computer vision. This is a course that covers um deep learning architectures with a focus on learning end-to-end computer vision models. Um, and so you'll be going over CNN's, VITs, um, V vision and language models kind of intertwining with one another, multimodal models, um, and proficiency in Python along with some linear algebra, calculus is, um, encouraged and definitely having some awareness about some of the, uh, probability and notation is is also important. Um, this is a a course that has four assignments and it's best for learners with a solid math and coding skills who just want to get a general uh lay of the land of what um exists in the computer vision um field. Next we have XCS224R deep reinforcement learning. This is great for learners who want to get familiar with ML to practice like deep reinforcement learning methods across robotics, visual navigation and control applications. Uh some of the topics include imitation learning, model free and modelbased deep RL methods and offline and online RL methods. Uh you have a you get a really solid um u comprehension of the basic RL algorithms that exist in the in the um field of of reinforcement learning. Um along with that um some prerexs to keep in mind is that a solid foundation ML is important. probability theory, multivariable calculus, and basic linear algebra is a mustave and uh experience with PyTorch is encouraged. At the time of filming this, um the assignments are still being developed, but you can look for updates um and check the syllabus listed under the resources doc linked at the end of the video. So, now we're going to talk about how to create your individualized path. Um there are some individual considerations to take into uh um account here as you're designing your how you want to take these courses to complete the AI program. Some of them include personal or career goals, professional background, time commitment, and your budget. Here are the course offerings for the AI program group by different topics. So we have the topics around NLP, computer vision, robotics, generative AI just to name a few. But if you are looking to figure out what courses would best align with what you're trying to learn or your or your specialties, u this is a great slide to figure out what courses you should be looking at. First, we'll be talking about our classical ML pathway. This is a path that is perfect for those who want to strengthen their AIM ML fundamentals and then branch into different specialties. So you can start off with our AI fundamentals course and then our ML fundamentals course and after that take any kind of track of courses that you wish. Um in this case you can do NLP afterwards or robotics based uh classes or vision classes just to name a few. Next is our NLP pathway. This is a pathway that tries to broaden and focus your um knowledge around natural language processing techniques. Um for a lot of these pathways, we always encourage you uh uh our learners to take the XCS221 course, which is our AI fundamentals course. This just gives you an idea of how our courses are laid out um for each assignment. So you'll get used to h writing proofs and also writing some of the code and testing things with the local and um local autograder that we provide for the assignments and once you get more familiarity with the course framework that we have um you can dive into our deep learning NLP with deep learning course and then from there branch into other specialties that you care about that will um definitely supplement some of the NLP knowledge. So from there you can it's recommended to take either deep generated models, machine learning or machine learning with graphs or even the computer vision course to to see how you can tie in your expertise of uh natural language processing techniques to different um fields. Next we have our robotics pathway. This is a pathway that specializes in trying to understand many of the algorithms powering robotics. And so we always recommend to try to start off by strengthening your AIM ML fundamentals. So XCS221, XCS229 are there and encouraged. But if you feel like those are really ingrained in you already, you can jump into the deep reinforcement learning course uh 224R and also our reinforcement learning course XCS234 which is a lot more applied to figure out what what algorithms exist out there for robotics and um controlbased planning. um um uh problems. And on top of that, um after you've taken some of those courses, you can also supplement your robotics knowledge with with graphs and vision. Um this is a great combo if you in case you want to have your robots be able to see, navigate, and make sense of the world visually along with um how to maybe construct um graph algorithms to be able to move your robots in more interesting ways. Next we have the computer vision pathway. The computer vision pathway is focused on vision- based applications. Um again we recommend starting with XCS221. Um but if you feel like you have the AI background already, you can jump directly into the XCS uh 231N deep learning for computer vision course. And from there you can branch off into the generative model course or the machine learning course. We recommend the generative model course after the vision course because there are some applications of vision in the generative modeling courses namely diffusion models to be able to generate images based off text for example. Um and on top of that um after you've supplemented more of your machine learning knowledge you can take some machine learning with graphs which would come in handy with some of the vision algorithms discussed uh in XCS231N. So here I want to preface that we've ranked our courses by the level of rigor. Rigor here means how challenging some of the courses are based on just feedback we've received from previous learners in the past. At the top we'll have our XCS234 course. It's a very theoretical and coding heavy course uh where the assignments are uh filled with a lot of uh learning opportunities um but it does take a lot more time. So if your schedule is definitely more constrained, we recommend maybe looking more in the middle of this graph. Well, you'll see that a lot of the courses that have around the same level of rigor are those from the different fields or the different um categories that um groupings that our professional program offers. So you can try NLP, the graph course, our computer vision course or generative modeling course. Roughly those are all around the same. Um whereas at the very bottom we have our AI fundamentals course. This is a really fun course but it allows you to still ramp up and understand some of the basic AI algorithms that exist. Next we have courses ranked by most applied to most theoretical. Um in this case we have at the very top uh machine learning with graphs our 224W course. The collab notebooks are meant to be very applied. that deal with industry level data sets and you can really take the collabs and then bud them into an industry application of your choice. And as we move our way down, you'll notice that some of the harder courses that we saw in the previous slides are also our most theoretical courses. So 229 machine learning is one of our most theoretical courses. There's a lot of proofs that you're going to have to write for the assignments in this class. XCS234 as well, but there's also a coding portion to that. And somewhere along the middle of this graph, we have courses that are balanced in the the application and theor theory. So in this case, um a lot of the other courses like NLP, computer vision, DRL, and um some of the u generative modeling courses all share around the same level of theory and coding. So for for these courses, you'll be working a lot on proofs, but then you'll also be working a lot on the assignments in a way that's feels more balanced. Next is a pathway for those who have um a background in math, stats, and theory. Uh we recommend starting with our 229 course as it is our most theoretical, but you'll definitely feel a lot more comfortable writing proofs and being able to um decompose different ML algorithms in a mathematical way. Um and on top of that you can slowly transition into practicing more of the coding and deep learning frameworks like PyTorch and NumPy uh by taking XCS221 first to get more coin with numpy and then anything else after that is going to be practicing a lot of the other deep learning frameworks like PyTorch for example. So here for computer vision, the deep RL machine learning with graphs, our deep generative models course and our NLP with deep learning course, they all handle building out much of the assignment algorithms with PyTorch. So there's some theory involved. So you'll be very comfortable at that. But for the coding based problems, you'll be getting a lot more exposure to these new frameworks. Next, we have the pathway for those who have a really strong coding background. So for those who want to dive deep into uh what AI and ML has to offer but and they consider themselves to be pretty savvy with coding, we recommend starting with our um AI fundamentals course here. You'll be able to definitely nail a lot of the NumPy algorithms that you'll be building from scratch um and then seamlessly kind of blend your way through all the different course offerings. So you can then transition to NLP, graphs, generative modeling or vision courses where you'll be focusing more on uh the heavier coding problems uh where there's a lot more code to write but it's using a PyTorch framework. So you'll be kind of navigating a new set of libraries to work with and then slowly you can move and and um increase the level of theory that you want in your ML pathway. So if you feel like you've had enough coding but you want to kind of get deeper into how to deconstruct algorithms from a mathematical statistical um sense then you can dive deeper into some of the RL um/rootic offerings and um along with our machine learning course as well. So here we've gathered a bundle of resources that could help you get a better picture of what to expect. You can find it in the link uh to the description of this video. Um, and if learners are curious about the assignments, you can look at the syllabi and or if you're more curious about the level rigor, you can watch and skim some of the content um, and through the grad assignments in the in the free resources section. If you want a general overview, you can check out the brochure. And if you want um, general questions about the program, definitely check out the AI Prod uh, FAQ. Thanks for listening to how to build your personalized pathway within our AI professional program. Until next time.