Cisco Director of Artificial Intelligence
Stanford University Artificial Intelligence Graduate Certificate, Computer Science
Current Time 0:00
/
Duration Time -:-
Progress: NaN%

How did you get to where you are today? What is your story? What incidents and experiences shaped your career path?

Summarized By: Jeff Musk on Wed Dec 18 2019
I can go all the way back to maybe my first program, I remember the first time I ever had a keyboard and started coding, was maybe at the age of seven or eight was back in the day using Basic, and I just fell in love. I had this book and by using that, I could affect Quiz Game as an introduction. And I wrote the first one and after that, like, you know, started coding over and over, like the same things. Then I stopped coz there was no internet. My family's mainly in the medical field. So I was thinking of one day I'm gonna be a doctor one day. But then I fell in love again with math and started looking into the engineering track instead of like the medical track in high school and focused more on that. And then I went to the American University in Cairo, where I majored in computer science and one of the stories that you always did like folks back then in 2002, when I was just joining the university as a freshman. First week with a friend, or begin a friend amid a stranger asked me like "what you wanna do it after you graduate?" And I said, "I'm gonna go work for Microsoft in U. S". And he was like, "Are you crazy that nobody's hiring from this university for various reasons back then." But the idea was like, that was absolutely impossible. Fast forward to three years I get an e-mail from our career advisement office, and I basically applied for Microsoft. You know, they came back to hire from Egypt, and that's where I'm from. So after going through all the series of crazy filtering interviews on the phone and in-person interviews, they offered work with Microsoft. After graduation, I went to the visual studio team, and I spent a couple of years there and then moved to the Bay Area to work on Hopping Lacombe back then. Then I got the bug of machine learning. I can't remember the exact moment where I saw the Qatar or some solution that is so elegant. You know, this solid data is the new code and how you can use data today. So some of the problems we've been facing that a lot of old systems kind of actually find a good solution. And I started tinkering around them, playing around with some of the problems. We had a work, and it was obvious to me that this is my passion. I'm gonna follow it. So I started working on some of my code decision processes for Improving the bear strategies for machine and what not? And then I just noticed that I want more. I want way more. So I, you know, got into the anti-spam team. I lit a couple of teams working on any of the little ability may transfer agents between like, you know that Jim Miller camp of your common Yeah. Come on, hold it, Players. And you basically busy owning ATA And you know, you have, like, filter spam different stages of email into the variance alarm. So I got into that and did it a bit more. But I wanted more. So I went to them where my team was either the team they're working on infrastructure and some modeling for the personalization and recommendations that you get on any newsfeed I think that's including the home page, whether you're on your computer or in your phone and this is where I really learned a lot. I was surrounded by people who are entire day-to-day is just getting to the best, performance metrics and assistant performance, serving machine learning systems and understanding user behavior and the modeling. I learned that time. We were had a couple of friends, we had an idea and we just wanted to go and do something about it. And it was, how can we use augmented intelligence, AI to more productive at work. We started with co-founders, we started looking at the meetings as a starting point for where we're actually spending a lot of time. And how can we automate all the desks? So you turn talk into action and you can start thinking of like, kind of bring the meeting into the enterprise workflow. So you are now our conference. Our agents could have easily joined this, recorded the whole thing, captured those action items and other tasks, but also give us our conditions over chat like, Hey, I heard you say this. What? You'd like to schedule the meeting next week, and if you say yes, it will just scheduled the meeting for you. So really integrating into like the enterprise book for what is lost in meetings and bringing that to the ecosystem that you're working on, whether you're using, like a desk manager or the calendar or any other system like CIM So bridging the gap between what happens inside the meeting room or in the conference line and what you actually use to work. So we got a quite bright cisco a few months ago and now I'm running the AI for contact center businesses and context interest solutions. Cisco is a big player in that space and there's a lot to be done here. We're implementing the agents, making them more productive and making supervisors more productive, helping customers who also call on the phone or chat or, you know, interact to get support and other, aspects of the context center. So there's a lot AI can do there, and we're very excited, actually building that here.

What are the responsibilities and decisions that you handle at work? Discuss weekly hours you spend in the office, for work travel, and working from home.

Based on experience at: Director of Artificial Intelligence, Cisco
Summarized By: Jeff Musk on Wed Dec 18 2019
I will actually answer the second part first. we at Cisco have a very cool culture of work and happen anywhere. We believe in the collaboration tools that we built. And most of the time, we're actually are working with teams across the globe across the US, across multiple places. So they can work from their home, office or while traveling. They work remotely. That's a big cultural value we have here that you can do work wherever you want cause we believe in the kind of like seamlessness. The main responsibilities when you think of what I can bring to our business is really about the direction and the way you introduce, change management and figuring out the technical strategy. So mainly, you know, if you think of where is it that you would have an impact on the lives and who are you helping? And seeing across the spectrum. How can you actually introduce AI and where would it mean that you know what customers need. And of course, coming up with innovative solutions for these areas. So for us, if you think about it, there's a lot of trends that are happening. As you know, that deep learning has been negatively taking. Should running like, uh, solutions into a direction where things like why more data gonna be more effective And you're solving way more problems that were or in a more efficient way. And then more on the Heidi effective ways than ever before and we're looking at, like, across the board. Or, you know, from the moment a call comes in or any interaction is initiated by the customer or even, you know, maybe before that as well. But I'm just gonna go through the, like, the path that you can see that for yourself going through. You have a problem where you're calling, banks or your airlines for something. How can we make your experience over time a great experience? What if we could just get you the best person to talk to right away. We can solve the problem or the team to talk right away for the concern problem. How can we recommend the agent to actually give context that could say you're browsing something on the website and you saw a product and you want to ask a question about the product? How can you escalate from the chat while preserving the context and find the right person who can help you with that? There's more that should be done. So one of the suffering we always looking to like, How can we take the call and make sure that we have analytics that would have you so problems fast? So there's a lot of knowledge that comes into a contact center, you know, the product catalogs and like them, a lot of knowledge. It has to do with the Catalogs and documents. Maybe they're searchable maybe they're not. I brought the ideas like how to expose this on the fly to the agents so they can be like superhuman right, that you don't have to go on like hold and neck. Put the caller on hold and figure out how can I solve this problem? That might be AI like psychics. That gives you the answer right away and just give that answer to the caller immediately. So there's a lot to be done on. As I said that before, the idea is to figure out the direction we're to invest, what kind of technologies we will use working off, you know, model into needs. And also we are aspiring to figure out the new one novel ways of solving problems. So it's not just the applied piece of it. There's a lot to be done as well. 

What tools (software programs, frameworks, models, algorithms, languages) do you use at work? Do you prefer certain tools more than the others? Why?

Based on experience at: Director of Artificial Intelligence, Cisco
Summarized By: Jeff Musk on Wed Dec 18 2019
This is a big question. I have been asked this question multiple times in different you know, situations whether it is software engineering, tools or programming languages or what kind of computer you use. There's always a question about choices. I believe in a couple of things. First one, there's no silver bullet, right? So there's no ultimate, perfect winner. There's always a situation of thing that you have to consider all the factors to figure out what works for me. So I continue just as an analogy. People ask online forums. Hey, should I use Go or Java? And the answer should be like for what? This a necessity or maybe a fit, you know, test first. If you want to have something that satisfies these preconditions or criteria, then you use this one versus that. So you have to figure out that first, what are you optimizing for right? But let's go back to the second point. I believe, more in a guiding decision-making framework rather than the choices themselves. Meaning that I would, you can figure out with the team what are the principles on top of which we can make decisions versus making the decisions because, in the future, new decisions or decision-making points are gonna emerge. And you really want to have a framework of resolving any conflicts or any competing interests. So let's say, for example, one of the decisions or decision-making framework guidelines we look at it like, Is it secure? Right. So if you're using a program, is it secure as prospective privacy of the customers you're collecting data. Most of the cases, that's true. So if you're using that program, like our framework TensorFlow versus PyTorch yeah, we know that they're coming from a reputable like projects, and it's not like a big or short. There's no issue there to use one of the two. But the idea is that you come up with all of these different guidelines or principals kind of ways of judging the decision making versus the actual process. I can walk you through a couple just thinking or given example. So security is top of my mind usually, especially for software engineering generally speaking, not just machine learning. The second one is gonna be something that will allow me an iteration velocity that will get us somewhere fast first, just to deliver something as MVP, minimal viable product. So, really, how can How fast can I reiterate this thing? If it's gonna be something, take a lot of time to iterate using the tool, then that is gonna be something that will slow me down every single time I need to use it. So you want to use something over bait evidence and accelerating your research or escalating your developing over time, right? So after that, you should think of all of the different criteria. Like does it skate? You know, does it save me like money and cost, like basically efficiency and so on? So these are just a couple of examples. And they change, depending on whether this is a short term thing or a long term thing to the other factors in both, I hope this gives you an answer. I can tell you about the specifics of what we used, but they're the choices themselves are circumstantial. And I don't believe in the silver bullet so I hope that is a better answer than just giving the exact concrete names and brands that people usually want to know. But I'm happy also to go into the details of what we have been using.

What things do you like about your job? Were there any pleasant surprises?

Based on experience at: Director of Artificial Intelligence, Cisco
Summarized By: Jeff Musk on Wed Dec 18 2019
That's a good question to ask. I think of philosophy actually, that is a reason for existence. Or like the French would say La Methode, in Japanese it is called Ikigai And that's kind of the overlap of four circles. If you can draw a diagram, it usually helps it presented. The first one is passion and as I mentioned before, explored enough to actually know where I'm passionate about, where I can spend like effort without feeling like you know, I'm hungry or working hard. It's just what drives me, right? So that passion is one of the circles of Ikigai. The 2nd one is what you're good at or competence gives you this push that you can actually do stuff without having to stress out a lot about them. So there's some foundation that you can build on top off and can know, I can do this because I've done this before or I have experience you to get me more mileage to actually learn this new thing and the newer thing after that and so on, and usually people would look at this as the flows on. It's not so hard that it's really impossible and stressful and a curating a wall, and it's not too easy. That's boring and mundane and maybe a bit repetitive. So this is the second serve of competency that you're good at. The third one, this really be the goodness fortunate doing it. Are you adding to the world in a more productive way? Are you doing good for the world? And I think most companies, definitely in Cisco, we'll check that box and then the last one is the compensation of, like how you rewarded financially, or other types of recognition. And if you can maximize the overlap of these four, you will get that kind of job satisfaction. Right? So what I like about my job a lot is that it's gonna fix all of these boxes and satisfy what I'm looking for. And it also gives me a lot off like ways of thinking about it. Hey, we have a lot of strengths here to deal with our problems. We will. You know, smart people have a lot of technologies. We can bring it to this that new AI approach to the product. So there's so much in use, right? Including that way, actually, like think affection of the three things that require or AI requires to grow. You need data, you need people who are developed and they have done the work with data and you need some accelerators who move faster and in like a multi experiment and you can see that we have covered all three here. So these are the things I do you usually look for and I think these are three strong points of working in a big company.

What are the job titles of people you routinely work with inside and outside of your organization? What approaches do you find to be effective in working with them?

Based on experience at: Director of Artificial Intelligence, Cisco
Summarized By: Jeff Musk on Wed Dec 18 2019
A lot of the inter-disciplinary efforts are extremely important. More than ever before, I think there's a huge need to actually work across like, if you take a cross-section of like approach, I work a lot with the product for sure. So the product managers understanding you know what is required to be done and why? Getting the guests And some think, the market artist week old sick when did field so the few vehicles, or, like the marketing sales, support the voice of the customer? I work a lot with the customers themselves what, you know, like sometimes like you want this like facetime. Take me back to the relationship to get an understanding of what is needed. So if you think of the other side, of course, internally, you have to work with engineers and your team and that you know UX is extremely important for me and for AI. There's something to be said about how to introduce novel techniques in automation, including the nondeterministic, AI solutions into products that have not been familiar with therefore comfortable with that. So understanding what the human-computer interaction is gonna be for your solution is an extremely important piece. And I work a lot with it. You ex researchers and New York's designers on that. So from savings marketing, engineering product you extremely important to get a kind of very comprehensive holistic view of what are you building? Why are you building? For whom are you building this? And, I think that approaches that I find effective and working with across the board is having a shared objective having a shared, common ground. Once you can motivate everybody to follow the same mission and pick the values that you put out there like we want agents to be more productive, we want to fix some of the issues that customers have been seeing. We want to do certain innovations that other competitors gonna do because we have all of these students here and so on. These are the things that drive people to actually work towards this one goal. Even though we have different disciplines, different backgrounds and different ways of the work. 

What major challenges do you face in your job and how do you handle them? Can you discuss a few accomplishments?

Based on experience at: Director of Artificial Intelligence, Cisco
Summarized By: Jeff Musk on Wed Dec 18 2019
I've been only Francisco for three months. I was just used up with the economy more backward-looking window. In any job that includes like Machine learning the fitness of data and finding the right data and cleaning up the data and pre-processing the data in a way that's easily digestible by the algorithms is usually something that I spend the look at some major time on. And my team mostly spends a lot of time on understanding. Like, you know, you're very shift after you really something and like the fitness overtime. So there is something to be said about, like how we have figured out, for the most part, how software should be released and how software should be managed. So the idea that the software lifecycle is almost scientific, been studied and figured out for a long time for many, many decades, right? We have some understanding to get back to a mature product in a kind of standard way. I think it's important to figure out something similar for machine learning, and this whole way of getting the data and shipping a product and monitoring the product over time. It's gonna be a huge thing that most companies now are looking at. We're trying to figure out what R&D standards we should be following for this kind of like solution or implementation of AI solutions. Some of the accomplishments we have. I think you know the secret of a really good model at any of the companies that worked at is finding the right data. So, really, dig deeper and figure out how would my dataset fit the problem and I had great success is because of that. Anytime you have like a leap, the magics, it's usually because we looked really hard at one of the three factors, right? It'll be like computers, usually there, because we can get more off. But it's the algorithms who have been using or the data, and more often than you know that your instincts, it's the data like, how can we get that idea for the problem, What techniques can be used on top of that data to get to equip, and that's a really good way of thinking. 

What are the recent developments in the field? How significant are these improvements over past work? What are their implications for future research & industry applications, if any?

Based on experience at: Director of Artificial Intelligence, Cisco
Summarized By: Jeff Musk on Wed Dec 18 2019
That's a good question. I mean, so we're in the industry, so you might get like some in your case, you might. This is old news with, you know, for us. It's important to see that what already effective things that's started from getting enforcing supplies from the industry as well, from being in search of other solutions. And really, they make a huge difference. So I can say, like, you know, let's go back in a few bits in time And the committee used the speech recognition field as an example. So for speech admission, it's been always that holmes always like, oh, processing of continuous data sequence data, right? So we'll use some set up some sort of a sequence model, Whether it's, you know, pick your favorite art. And then one of my models were based on, like a Bilstm, like, you know, understand the can strike the audio. But then we saw the hole that, you know, shift in multiple, you know, fields on just speech with the attention base models and transform models. And we saw that yet that also helped us a lot and improving the speech mission task. So the applications for us is that we like to see multiple, you know, fields or multiple focusing proof, right? So whether it's in the efficiency aspect of like, how fast can I do training and experimentation? How fast can I do it in French? How cheap can this be because at scale, when you're actually thinking of, like hundreds of millions of users, right, you have to say, figure out that every cent counts, right? It fewer are doing something. Committed improvements. These are extremely important. The same thing applies for figuring out the metrics improvements. So if you can use a new technique or a monster or a data set, you can get on improvements in and the metrics then this also leads to the customer experience being improved a lot, and this helps your business grow. So these are usually implications. I think that there should be some sort of like a look back, and I think there's a lot of base every time you know writing today, we can get the industry into the research and the research help, you know, have improved these two aspects of three aspects in the industry and so on you know, a virtuous circle. For us, I feel like, it's gonna be more than speech before that's in speech and for those recommendations. The new techniques that come in can help it accelerate and reduce the cost of AI Solutions.

What qualities does your team look for while hiring? How does your team interview candidates?

Based on experience at: Director of Artificial Intelligence, Cisco
Summarized By: Jeff Musk on Wed Dec 18 2019
You know, there's something non-technical aspects. And there are some technical aspects, right? For the technical aspects will usually look for, like, you know, intellectual horsepower. Given that, new problems emerge, how can you deal with, ambiguity and challenges? So for me, we wanted in question is usually like, okay, give me the hardest thing on for breaking down the understand, like, how did you achieve what you achieved and what challenges you faced and who have really big these sources. So it's more about that, you know, the way of thinking the intellectual horsepower. And I personally like the appetite and the passion for learning because I'm looking for where this person is gonna be in the future. And I like this kind of exponential orient around me in your growth at multiplier value also for the team and that you have an ethic, you know, way of helping other knows shifting at it. The more words did not technical aspects. I like the multipliers effect like do you like to help others is a huge thing. Cooperation is a huge thing for us, and it's also the nature of having to research and, you know, like science. And then she running. It's a team effort. Usually, the cost pollination off ideas can help a lot, especially from different fields, Right? Some ideas. You can see that, you know, being the cost pollinated for my vision and speech and so on. The other aspect is we look for fit of the current team culture and the organization and the company, their values we respect and, trying to figure out, how good a fit this person who is coming in and gonna head the culture thrive more and the depends on the kind of values that you cherish and the ones that can make the fabric of your team strong and I think this is an important thing when you see like somebody coming in, how does the team feel about thinking of interactions with them and collaborating with them and basically our understanding is the kind of values, your respect or how you admire your team in your organization. 

What are different entry-level jobs and subsequent job pathways that can lead students to a position such as yours?

Based on experience at: Director of Artificial Intelligence, Cisco
Summarized By: Jeff Musk on Wed Dec 18 2019
There's no shortage of paths, especially today with the high demand and sheer learning in the eye. I've seen students from all the different backgrounds of chemistry, physics, mechanical engineering and so on. Today that students who are going through entry-level jobs but maybe came from a different background. They are very strong in statistics, the very strong and like, you know, modeling. But maybe not as strong by virtue of not being trained as a computer scientist, to can close that gap between the people who can build it serious systems and large systems of companies like Netflix, Microsoft, Cisco and so on. And once for me, it came from like a different background on there. Now they found those that themselves into this entry-level or maybe a couple of years in at the company where they have to interact with such big system stated assistance. So the book is called Compuing with data, Computingwithdata.com is the website and you can actually see the examples and understand the skills it acquired right to go there make it. I think that would sum it up more than anything else. We have it in writing there. So the idea is that the background will actually maybe matter in the first couple of years. But after time you will see that confusion of skills and you wanna build the skill set that will fit in overtime, where your trajectory is headed. So said the target and see that okay, who else in this field has made it that level? What kind of accomplishments they have made and what can I think from the steps he took. And I think the Internet is also very important that you figured out a mentor who has done this before and we will guide you through the way. 

What were the responsibilities and decisions that you handled at work? Discuss weekly hours you spent in the office, for work travel, and working from home.

Based on experience at: Software Engineering Manager, LinkedIn
Summarized By: Jeff Musk on Wed Dec 18 2019
Similar to what I'm doing now with maybe on a more limited scope. So my team was the news-feed team infrastructure and like some of the folks I work with also did modeling they had responsibilities across the board from, like figuring out the future extractors that scale, remember that back then, right. And then we had about 400 million users on us, maybe 600 or more. So I'm not gonna be surprised if I hear 800 anything. So the idea is like, this is an ultra-scale break. It's an unbelievable the amount of data that you're ingesting and extracting features. And that building is like high inescapable systems to know like you know what to do a turn time and all this is public, actually. Go to the engineering block and see how LinkedIn builds, the newsfeed systems. And this is that speed and the approach, you know, that the feed team took was basically way have, like, a middle where that can query multiple sources that will be ours. And it's like the first passed rankers and they have their own that, you know, kind of updates that you wanna put on top after that here's a lender or something that ranks which ones should be shown at what time? I remember, like, one of the first projects I started as an individual contributor, like a couple of months. Six months, and then one of the projects I worked on and my team then took over is that you know, Look, when you see something on the news feed, you won't remember this hard. But how you do this for billions of updates today, right? They literally it's so much going on there, it's got them to send the scope off like, the complexity of the systems first and then people in something that works at that scale so for the team and that job responsibilities were, like again guiding the team, like okay, with which ways we should go have products on approaches. We should take my word being that scale and also modeling. So some hooks on the team. Burdick, you did great, you know, molding the accuser behavior and figure out a new features. New ways of introducing, you know, new kind of updated or new techniques to the defeat newsfeed models and understanding the impact of ow did this impact these matrix that are coming up with the right matrix is also a challenge and so on. But if you think about that, to sum it up, it's also kind of similar kind of requirement we're dealing with which somehow something which more affects future in engineering. So representation and optimization at times also, you know, improvements. And it did. Modeling and evaluations. It's pretty close to what you know, all the basic AI projects, including the famous tech article about, a few things to know about machine learning. So we usually think of it in these three axes, as the team has improved along with the experimentation, velocity. And I'm standing the history of like what people have done before and trying to understand that the differences between experience back then and experience now they're very busy off the data sets that we have our features doing the extraction at scale and sometimes the pre-processing and all suppose processing the results like feedback loop for, continuous training continues evaluation so on. So there's a lot that in data in one scientist. But I think overall it just to give you idea hours spend at the office. I think most of the jobs we do are like, you know, more than eight hours speaking out, very candid about that. For LinkedIn, there wasn't much travel. But I think it's also fixing your environment where you can work from home. It's a very good culture. A lot of important things still go back there like I think once a month is something called Indays. Basically, When you don't actually do actual work, but you go out into something, maybe with the team or to some activities that you can go on like network with the team on other teams around you and each one is a theme. So it was a good cultural thing. I think that.

What was the hiring process like for your job? What were the roles of people who interviewed you? What kind of questions were asked?

Based on experience at: Software Engineering Manager, LinkedIn
Summarized By: Jeff Musk on Wed Dec 18 2019
For LinkedIn and this might be updated now, but the idea is that you get a job description and you can feel this kind of experience the company is looking for. LinkedIn hires for the entire company. So there's a committee, you know, a panel to interview you with the entire day. Just a standard thing, but the idea is it has to happen from all across the company. And, for some engineers, you might get to know certain people who have been working on something similar, and they differently kind of close enough. And the experience, bit higher and they can expect a certain kind of interview it's like design, coding. There's one with the hiring manager that is a kind of culture fit and understanding. Like, where you are coming from or what kind of things you've done and there's one for appropriate communication as well which is I think, very important explain things in ways that can facilitate collaboration and improve the way people you know, built like solutions. Removing ambiguity, or, just like seeing that how we communicate overall. And, the questions asked. I can't really talk about them, but you can expect, like the standard questions, like design this solve this problem and so on and the roles of people who interviewed me, back then this like a distinguished engineer who had the launch interview, had a very good conversation I like that kind of thing. Somebody who has a very broad understanding of the history of the company and people are like in the same experience level. Questions at that level where you actually have to solve problems for the ultra-scale turning you have to deal with

How did the program prepare you for your career? Think about faculty, resources, alumni, exposure & networking. What were the best parts?

Based on experience at: Artificial Intelligence Graduate Certificate, Computer Science, Stanford University
Summarized By: Jeff Musk on Wed Dec 18 2019
So I think this is a very smart move from educational institutions that now we can go and you say, I just don't want a degree. I want to learn, right? And I think this has to do with the explosion we saw in the books and how people are gaining information and learning these days. So the program at Stanford is open for professionals. You can apply online CPD and then you can fill out an application. In this case, there's a process to get admitted. And then you have this catalog off like really awesome courses that you can pick and chose from, I think for the artificial intelligence certificate, the requirements are extremely like, you know, focused words like How can I be successful at an AI company or my career? There's one acquired class, which is CS221. It's kind of a survey class going through all the AI like, like did you sense you could, you know, past or see a different aspects, maybe like an hour in the industry and concealing it goes through like castigation and digression all the mission in classical ways and some of the Deep learning aspects and Marco decision processes and graphical models. And then some of the, symbolic, you know, on the first-order logic. There's like a tasting menu of a lot of things, and it's a really exciting class and then some of the stuff that you can learn a more classical but focused mission CS221. You can understand that okay, what does it mean to be, like a complex model with some learning theory principles? And how much did that I need? Like the bounce on that? The relationship between the train together in this error, what's the VC dimension and all these things that you might not see it in the industry But here it is. I think it's a strong foundation that you can learn from, and then the class. Actually I'm a teaching assistant for it now is the CS authority or deep learning. And that's I think, all the hype here a department now is extremely important because that's where a lot of the industry, direction and industry, companies and teams are moving towards like, hey, I wanted small, but I have this amount of data. How can I make use of the data in a way that representation actually gets to the solution I tried to achieve and it takes you through, understanding DeepLearning. Why does it, solve some of the problems like others can't solve and at the importance of that building your own. And then it goes a little bit higher level and can start using it frameworks like TensorFlow and PyTorch and can build like your own single models. Then talks a lot more about that machine learning in practice. So that's the part which I think is unique, very important. This is the third time I mentioned in this interview, and I think it's really important to know the life cycle of your machine learning solution or AI solution. And you know there's some concentrations or modules that are focused on certain areas like CNN's and Convolutions and bookmark computer visions. And then there's also a module that's very focused on sequence models. That course has a very strong emphasis on the projects, end to end projects. You know the quarter system is 10 weeks, project is of eight weeks and makes 40% of the grade, very important cause it tunes in the next everything you do in the industry research. And it has the entire flow, like the life cycle again looking at the data problem, input, output and their multiple things you have to go through to actually achieve a solution at the end. So it was a great experience. I think the best parts are going through these projects especially for like two to nine and to thirty but also understand foundations, the math, why things work and getting a broad perspective about how different techniques can be used in different solutions in different problem spaces.