How to Build a Real-World AI Project Portfolio

Sam hello everyone. We are now live. So welcome to how to Build a Real World AI Project Portfolio. My name is Rosanna de Rivera Gomez and I will be your host and moderator today. I am a Senior Data Scientist at HACOM Time Series and also a Product lead at Omdena for those who are new to Omdena. Omdena is a platform that puts the main experts, experts and AI professionals together to solve problems related to social impact and I've been working there for quite a long time and also as a mentor in many many roles and then transitioned into industry in the energy sector. So it's quite nice to be here today. We appreciate you taking the time to be here with us to this event and we are also glad to see you joining from all over the world. Today we have attendees from 140 countries so that's really a lot. If you happen to be awake and it's late for you. Thanks for being with us. If you happen to be awake, it is early for you also so let's get started. So our event today focuses on building an AI portfolio. So for all those of you who are interested in a career in AI and wants to know how to build your skills, how to practice your skills is a great place for you to be today. We are going to get right into it today. So if you hear anything in our discussion that is interesting for you that maybe you have a question about and you would like to ask our panelists, please follow the slido link that it should be seen in the description box on the YouTube page. And also if you're just listening and you want to you see some questions that may be interesting, you can also just go there and also upvote the questions. And at the end our speakers will do their best to answer all the most upvoted questions during the Q and A and as the portion of this event that is going to be towards the end. So without further ado let's get started with our speakers. So today we have Anna Koroleva who is a PhD in Natural Language Processing and is also a computation, linguistics and NLP data scientist. We also have Mohamed Amer who is long term Omdena collaborator and is also a machine learning engineer at a US based startup. And last but not least we have Kennedy Wangari who is an award winning data scientist and developer advocate. He's passionate about empowering AI technology in communities and especially in Africa and has been fighting for more inclusive and ethic AI throughout the world. So thank you all for being here today and welcome we would love to hear a bit more about you. So maybe if you want to give us a short introduction. Ana, would you like to go first?

Absolutely. Thank you for introduction Rosanna. It was great and it's my pleasure to be here. My name is Anna. My I have training in linguistics which is often considered a non technical domain. But despite that I work in AI more precisely in Natural Language processing or NLP and as Rosanna said I am very interested in biomedical NLP. I have a PhD and postdoc in that domain and I also have several years working experience in industry in text analytics. At the moment I live and work in the north of the UK in a text analytics company called Relative Insight and that was it for me.

Thank you, thank you and welcome again. Kennedy, if you would like to go next.

Great for me. From joining NCBA bank in 2018 where I served as a data science internship role, I've been able to grow through the years to currently working at UNEP as a technology consultant. Briefly my background is information technology and following my internship in 2018 I decided to devote myself to improve my data and literacy levels through Coursera and specifically through the deep learning specialization courses. I've been able to grow through Coursera community and as I mentioned that's why I'm passionate when it comes to learning and building with a community of learners and I've been doing this through the deeplearning AI community in Nairobi and within Africa and more to that. In line with today's session I've also been actively involved with Omdina having served as a machine learning engineer in a couple of projects and beyond that working on projects and learning through Coursera I was able to realize my passion in NLP and I've been able to join and work with the Masahani community basically helping build open source language data sets and I'm excited. Today's session on looking at how do we move from courses to our real world and building portfolio.

And thank you, thank you and welcome for being here today. And last but not least also Mohammed.

Thank you so much Rosanna. Hi everyone, my name is Mohammad Arm. I'm a machine learning engineer and data scientist from Egypt. My major study was electrical engineering, Then I studied computer and systems at Enchantion University. That ended up me with seven years of related engineering studies. On a professional level I joined several companies here in Egypt, Switzerland and also in US. I work remotely with iOS company and we are working on detecting and estimating the property damage using computer vision and virtual reality. Thank you so much.

Thank you all. So we have quite a diverse background here. I'm already very excited to hear more about your stories and the road that you had to get into the positions that you are today. So let's get started with our discussion then. Thank you for the introductions. Now we will have about 30 minutes of discussion and again the reminder to you if you're watching this, if you want to ask the questions to the panelists, you're welcome to follow the link that you have on YouTube and ask us our questions. And also I think that if you are talking about careers in AI and so on, I am expecting that we'll learn a lot, not only about the road that you had, but for sure a lot of advice as well. So quite exciting. So let's maybe start this conversation with a question about how you got into the position of where you are today. So a little bit if you want to share about your background about the road that led you into a career in AI. So whoever wants to go first, feel free to start. And also anyone else who would like to give some feedback or to just continue the discussion, you're welcome to jump in. Maybe Muhammad or. Or not.

Yeah, sure, I can start with that. So like many AI practitioners, I started with Andrew courses, the machine learning specialization and also the deep learning specialization in Coursera. Then I like tried to like test different resources like Super Data Science with Udemy. They were like the best selling there. Also the Samurai, Dusty nano degrees, all of them were like really wonderful. But the thing that really helped me a lot is reading more books about AI. The hands on machine learning experience. It was a really great resource of learning and also it's important to get some hands on experience. So I tried to apply what I learned in these courses and try to simulate the real case scenarios of AI and something that I can build with my hands.

Thanks. Anyone else had a maybe different path? We hear a lot about people starting with courses and then moving into projects. So I don't know if that was the case for everyone. Maybe Kenneth, you would like to yes.

When I talk of my journey I look at three communities that have been bivortal and as like I did earlier mention. Upon completion of my data science internship in 2018 at NCBA Bank, I enrolled for various courses at Coursera, specifically deep learning AI courses from the deep learning specialization and I think I've almost enrolled, taken up almost all the specializations and upon gaining relevant fundamental skills in 2019 when Omdina came on board, I joined the first. I've been able to grow through Omdina, that being the second community and that community is the overall deep learning AI community. They've come in handy in terms of from gaining the requisite skills in deep learning in machine learning in data science. And Omdin has come in handy in the various projects that I've been involved in to augment on my skills, to apply the skills that I've gained towards real world problems. And the global Coursera community has come in handy for me. So when I look at my journey through my data science internship at Nvidia to Zetova and now at working with unep, those three communities have been quite critical for me and I'm grateful for the work that they're doing towards supporting the community of learners.

Absolutely. It's quite exciting to see so many communities also like with so many opportunities for us to get experience. Right. And what about you Anna, how was your path into AI?

I think it was roughly similar to others, but just a bit more detail. I am a linguist, as I said, so we had some maths and programming at the university, but nowhere near enough for a proper career in AI. So at some point after my first job, which was in text analytics, I realized that I need to learn more about machine learning, about even just programming. So then I did my PhD and I learned a lot of things, but mainly I learned that there's so many more things to learn, which is I think a constant thing on the data science path. You just, every next thing you learn uncovers 10 more that you need to know. And I think it's quite maybe funny that actually I started, I started taking Coursera courses after my PhD because even at that point there were things that I didn't know that I needed to learn. So again I took one of deep learning AI specializations, the one about TensorFlow, because that was something I needed at my job at that time. And I think it's, it's really good and brilliant and among many, many courses that I've taken, it's one of the most comprehensive and it goes, theory goes hand in hand with practice, which is not, is not always the same at other courses and platforms. So I really like that one and I'm, I'm going to do more courses from deep learning AI because I think they're one of the best. Genuinely, this is not advertising, but I think it's also good to point out that when you do courses you find some more courses about the slightly different aspect or another technology or another library and you just keep doing that and at some point you feel stuck because you don't really know where to go next. And that's where Omdena came in for me. When I realized that, okay, I. I've done probably all the courses I could find on, let's say introductory machine learning, but I needed to apply it. And then I started doing projects with omdana, which are AI for good projects. And they're exciting because they are all about some good cause and they provide excellent learning opportunities. And I think that really helped me in many aspects, that helped me meet a lot of great people and uncover many opportunities in my life.

Thank you. That's really a lot of insights. And I think that one common thing that we all have, coming from a background, from another background and moving to an AI career also, because AI is kind of like a new, newish career, so it's very common that you have people from other backgrounds and we always have to learn new skills. I think what Anna said is a very true that we are always like in tech, we are always learning new things because the tools are always changing, right? So for you, like, we have this, this on, on one side you have to learn the skills, you have to learn how to use the tools for AI and at the same time you also have to apply them into projects. Like I always say that data science is very applied. So for you guys, like, how do you see, like, do you think that kind of like a question that I see a lot is like, when is the right time for you to. When is it that you are, you can feel that you are ready to start a project like, where you're kind of like beyond the phase of learning. We know that we are always learning, but that we can move on from just learning like the basics and then getting like, getting your hands dirty. This is something that I think that is very common. And also, like, if you want to share some mistakes on the road that I'm sure, like we all have and like the learnings that you had from that. So who would like to go first?

I could maybe. I have a brilliant failure story here to share and I really think it's very funny now, but when it happened, it was a painful one. So the story goes that I took one of the biggest, greatest courses on Coursera for machine learning. It was really good. It gave introduction to the maths behind it and many, many things that a lot of other courses are missing. And then I had an interview with quite a big company that was quite excited about and I just froze. I couldn't answer a simplest question. I had it all in my mind, but I just couldn't find the words to say it. And I think that provides an answer to your question. You can take a lot of projects, you can do a lot of things, but first it's all a matter of confidence. You can do Coursera courses or other learning courses and maybe you will ace interviews after that if you are confident enough. But if you're a bit like me, and if you're not, there is a chance that you will just freeze. And so after you have some basics, it's really time to go out in the world and start doing something. Like Omden is volunteering. So it's a great way of getting experience. And another thing that's great about it is that there's a lot of projects in different domains. For example, my passion is AI for healthcare. So that's if you can do courses about machine learning or nlp, but you will probably not find that many courses about AI for healthcare, you will need to go and find a place to apply those skills. And it's not so easy on your own because you will struggle finding data, you will struggle finding a relevant task that will be, you know, achievable, but also important enough to basically to convince anyone after that that you did something interesting. And Omden is really a place for that. So I would say of course you need some basic, basic courses because if you know nothing about Python, it will be a little bit difficult or if you know nothing about machine learning, you can learn then, you know then even without a lot of knowledge. But it's best to have the basics do some of the, you know, best known, most fundamental courses and then as soon as you can jump into real world projects.

Awesome. What about you guys, Mohamed, what do you think?

Yeah, okay, so I advocate for learning by doing. It's really important to be like an active learner and to apply everything you learn you haven't to do something big at least if you learned new concepts in machine learning, for example, classification, so you have to go directly to your console or coding environment and try to draft some codes to understand what you learned and gain the necessary experience for that. After you achieve like a certain amount of experience, then you can look for like some internships, like some volunteering projects to like bring your skills to the next level. For the other question you asked about the mistakes that we did during our journey, basically I did two mistakes in my like career journey. The first one is to that when I started my like my learning, my machine learning learning and like education, I was like aged to learn everything about machine learning, nlp, computer vision, reinforcement learning. Especially in reinforcement learning I like I studied like a deep, deep technical things that I don't use currently in my day to day job and I only use computer vision for my current job. So it's really important to you can take some basics from every domain and then to specialize on single domain and to gain in this area and have more experience. The second mistake that I did that when I started to do projects I didn't care too much about the domain knowledge that I work for and even if using the same technology, like for example my case computer vision. And then you started to switch between different industries, healthcare, agriculture, anything else, Smart cities for example. So every time you join a new industry you will join like a new learning journey. You'll be like totally new for this domain and it will like you'll lose time and also effort to switch between industries. So keep focused on the technology that you learn and also the industry you want to enter.

Perfect. Yeah, I couldn't agree more. Like this topic of like an infinite loop of courses is very recurrent and also it is a like a fun fact for me was also one of the things that I considered that was my mistake coming maybe a little bit like Anna coming from academia we sometimes are like a very, very expert in some specific area and then we think that we have to continue learning and we have to be like the person who knows everything about everything before you start doing something. And that's definitely not the case. So lots of learnings. Thanks for sharing Mohammed. What about you Kennedy? How was your like how do you see the this transition between learning and getting more like a hands on experience.

Great. What I've seen what worked for me was after gaining the fundamental skills in both Python programming, having come from an IT background and stepping up to the real world is that gaining starting off first to gain the fundamental skills be it in data science, in machine learning or those different topics comes in handy because this prepares you for the learning by practical approach. Because what I still work for bu is that having taken up a number of deep learning AI courses on Coursera and joining now omdina to work on a number of AI for good projects. It came in handy because I was now able to cement on the skills that I'd gained be it in machine learning, in data visualization, in the different topics in Python programming, whether it's web scraping. And also the advantage that came in easier for me was that I was also able now to gain hands on exposure and experience and build up on the skills that I gained from the courses. So I would advise someone to first gain the fundamental skills, be it in Python or R programming, in machine learning in data, and then step up to embrace the learning by doing approach that's now embracing on projects. You could start from beginner level projects, moving all the way to intermediate advance. And Dina comes in handy when it comes now to gaining the hands on approach and experience. And along the way I also realized that I was also able to build a portfolio of the projects that I was working on along the way. And I was also able to assess my progress because I could now tell myself at this point now I've been able to complete these, this level of particular project and I cannot be able to take up another project. So being able to fast gaining the fundamental skills and then stepping it up now to embrace fully the learning by doing approach on working on projects comes in handy. And for the mistakes that I committed along the way, I would say when I started taking up the courses, I took up a number of courses, like let's say maybe about five at the same time. And that became quite overwhelming and at times I could lead to burnout or maybe slow your progress. So what I came to learn is that be objective in the number of courses that you're either learning at a particular time so that you're also able to understand the various topics. And this came in quite a challenge. Like when I was starting off doing the deep learning specialization at the same time I'm doing data visualization specialization and other courses at the same time. So I think that was one key mistake that I learned from, from when I started taking up courses. And being objective, like maybe taking up two different courses at the same time while working on your project can really help you so that you don't get overwhelmed and at the same time able to cement your theoretical skills and be able to apply the skills in the various projects that you're working on.

Awesome. This is like really, really precious advice. I couldn't agree more. Also like this, this, like we have this loop of courses. I think a lot of people, especially the ones who are starting, sometimes it's even a bit hard to know which courses to take because the field of AI itself is so broad. We have like as Mohamed was saying before, we have computer vision, we have NLP, we have McKay's, also time series and so on. So it's difficult to learn which path you have to follow. And it's very tempting to actually try to just learn everything. Then that's definitely not the most focused and efficient way of learning. So I really like your advice Kennedy, about just try to take one step at a time and take projects this way. You said it's very cool to take projects that are like different levels of projects and you can take something that is totally data science but is much simpler than some very complicated advanced project. And this is also like an experience that you gain. So pretty cool. So moving on with our discussion since we were all touching upon this topic of courses and learning and so on skills, what would you guys say that were courses that you took that really help you to prepare you for your career journey and also especially like the skills that you gained from those courses that you apply nowadays into your current job? Who would like to go first?

I can go first. So as I mentioned earlier, earlier I talked like some specializations at Coursera. It was like really important for learning fundamentals of machine learning and deep learning. Then another courses that focus focuses more on hands on experience like super Data science courses on Udemy. Both of them were necessary for me to gain the fundamental knowledge and also hands on knowledge after that as a next step. Reading books is extremely important to bring your skills to the next level because courses like as it are not enough. You need to like have some knowledge in depth and you can find this knowledge just in books on this. What I feel yeah, great.

I think for me I have a number of specializations that have greatly molded and shaped me in my day to day work and my journey up to thus far. And one of them is the Mathematics for Mathematics for Machine Learning specialization from Imperial London College. And this specialization particularly introduced me to various topics in linear algebra, calculus and multivariate analysis. And sometime back last month I came to realize the great wealth of the skills that I gained when it came to debugging and scaling the models that had given us quite some challenges for the last two months. And the skills gained from that specialization really came in handy to even figure out where the challenges were and being able to scale and debug and have all models working. Another specialization that has been quite pivotal in my journey was the first specialization that I took in 2019, that's 2018 rather that's deep learning specialization. That introduced me to the various topics in deep learning and that said my journey towards understanding the various topics in deep learning and also how to build a career in deep learning and came in handy when I was doing my interviews with Nvidia because some of the questions touched on some of the topics tackled in one of the specialization courses that is structuring your machine learning projects. And lastly that specialization that has been quite pivotal is the NLP specialization because this particular specialization introduced me to the various topics in natural language processing, how the various topics in text classification, machine translation and other topics in nlp. And this has come in handy for me because working on one NLP related project at UMDINA made me realize and discover my passion in nlp. And this has informed my journey up to thus far where I'll now be joining at the University of Edinburgh next year to still focus on NLP with a special focus on Neuro machine translation for Kenyan languages. And more than that it's the skills gained from that specialization has come in handy on the work that we're doing with the Masakane community for the Kenyan languages. So for me I'll talk more and more on the many specializations that I've learned and gained skills from. But those three have been critical in my day to day work and critical in supporting my journey up to thus far. And I'll continue learning and also taking up more projects for the greater good.

Wonderful. Yeah. I think that one of the most beautiful things to me nowadays of our just like our experience at this time it's that we can really learn like a learning recovery, quality information from online and so on. So like and also in communities where we have people all over the world and we are much more connected and also connected through knowledge. So that's quite exciting. Exciting. And also under this line of in depth knowledge and specializations and you know like getting like very deep into your skills maybe I'll have a question maybe more to Anna or Kennedy. So you guys have a research background as well. So like if someone were interested in getting to deep learning by taking like what courses would you Recommend? Especially like a 4 deep learning AI. AI we know that there are many like a very, very advanced and but also very in depth courses. So did you take any of those courses and which ones do you recommend for someone who's more interested in going to this direction of research in deep learning? Anna, if you like to.

I can try to answer that. But I think again as I said I first did my PhD and then I came to Coursera and now I wish it happened in the other order because it just add so much more depth to what I was able to do during my PhD and it would make my life easier at the time. I probably won't be naming some particular courses or specializations because actually one of the things that I wanted to say for when you choose courses it's quite important to know that data science and AI keeps advancing. So one thing that I like about certain Coursera courses, for example deep learning AI, is that they don't stay, they're not fixed in time, they keep updating and changing things because AI changes. You can't make a course and have it relevant in five years, maybe not even in two years. So the courses that I've done, I constantly receive notifications from them saying that they've updated them, they added new things, they restructured them. And I think that's one of the things that is worth looking out for. If a course course is very old, it should be maintained, it should be updated. So one piece of advice that I feel like giving is look for well established organizations that you know that have a, that have a record in giving good courses because there's so many courses everywhere now, there's so many platforms. So just don't try to chase all of them. Find one provider to providers that you can trust and rely upon. But also don't just go for all this course in machine learning, it was probably great being the first one in the domain, but now it might not be relevant. Look for the courses that are updated and then of course if you want to get into the research, it's good to think of what exactly you want to do. So again, is it computer vision, is it nlp, is it in biomedical domain, is it, let's say conversational AI chatbot and try to focus on that because again, the field is huge. So by trying to chase every course you will spend a lot of fun time learning, but it might not get you closer to your dream, to your target in your career. So try to try different things, figure out what you like and then push for that.

Very good advice. I don't know if anyone else wants to add. If not, then let's dedicate our last minutes of this discussion into building portfolios. So building AI portfolios more specifically. So for all of you, I know that there are many different opinions with it. What Like I think that for someone who's starting and who's trying to build a portfolio, there are many questions and maybe the most obvious one is like what projects should be there? Like what is like a good enough project to be in your portfolio, like where do you stop? And so on and also like how many you should have and if you are interested in AI for good and so on. So like how you can find this and also like find fun projects or what would be a good idea that you're not Just making something for the sake of having a portfolio. But we're also like learning and enjoying the process like a little bit. What Ana was saying before, like enjoying the learning but also by doing so. I don't know. Mohamed, maybe you would like to.

Yes, sure. I did three things to create my AI portfolio. The first thing is to like just check quickly some competitions on Kaggle try to resolve like their like challenges there, try to have like higher accuracy etc. It was like really fun but it was like more focused on the like mobile research or how I get like improve my model architecture and all of the things. The second thing that I did is to like attend hackathons different like different hackathons that was more focused on machine learning. I joined more than 50 hackathons as I remember in Egypt and also virtually and every time I was trying to leverage machine learning to tackle different challenges with an end to end solution starting from the website or a platform that you're using and how we can integrate the machine learning back end to make it more functional. The third thing is to volunteer with on DNM to join like people from all over the world to tackle global challenges like finance, like any things related to NLP and other challenges that we are facing in our planet.

Cool. Kenti, would you like to go next?

Great. I look at the various steps that maybe worked for me when it came to building my portfolio and one thing along the way that I've come to realize is that there are two types of projects that we showcase in our portfolios. And I'm looking at the first type of project being the code based type of project in that you're building either dashboard or you either building an application that trying to solve a particular problem that you showcased in your portfolio. And these are the kind of projects that really focus on showcasing your technical ability, your ability to code. And the second type of projects that you can showcase in your portfolio that I've come to learn of is what we call the content based. In that you can also be able to showcase your various blog articles that outline how you solve a particular problem or maybe a YouTube tutorial of how you tackled a particular problem. And with that in mind, on the various types of projects, that's the code based types of projects and the content based types of projects. We ask ourselves what are the steps into how you would embark on building such kind of a portfolio and where would you host such kind of a portfolio? And what I've seen important out there or what's important is that or what I saw work for me is that one is look at the kind of projects that have a meaningful impact. And by that we mean the very many don't work on just any generic project. You're looking at having either an MNIST or a Titanic project. That's yes, those are good projects, yes when you're beginning off, but those wouldn't be really wouldn't send a good impression to your audience or the people who are interacting with your portfolio or the projects that you've worked on. And I've seen that being able to look at your local setting or your industry and coming up with ideas and looking at projects that you can be able to solve in a data driven way like Mohammed mentioned, maybe you can look at in the financial domain or in the healthcare domain and look at how you could solve that particular problem and then maybe showcase it in both a code based version and also a content based way, like having an article follow up for the same. And also I've also seen being able to engage in communities and look hearing their pain points, hearing the challenges that they have would also give you interesting ideas on the kind of projects that you can work on because you'll be able to listen to their problems and if you and the problems well, you can be able to attempt and work on those problems and solve them in a data driven way. And when it comes to how many projects you'd showcase in in your portfolio, I think for me four to five projects are quite ideal in that again you're looking at if you're beginning off your career, you probably maybe have a few companies that you wish you'd want to work for. And a good thing that has worked for me like worked for me when I was interviewing with Nvidia was to look at the original going projects with him and being able to tailor my portfolio of projects in alignment with that are related to their use cases and also so that the portfolio projects that you will have either in your resume or in the different platforms that you'll host them, they're aligned to that particular company or also aligned to that particular role that you're looking out for. And this showcases to the hiring manager or to the person that the kind of value you'd bring to that company. And once you've had the different ideas or the different projects you'd want to work on and maybe you have now an idea to the projects that are related to that particular job description, either data science role or machine learning role and you've tailored your projects in that either in that domain or in that particular for that particular company. We ask ourselves where can you host or where can you host your portfolio projects? And the go to platform has always been GitHub and that's one of the great platforms when it comes to hosting your projects. Another great platform comes in that's LinkedIn. It's a great platform to share and showcase your write ups or the various projects that you're working on. Twitter has also come in handy and other microblogging platforms you're looking at. Remember we also talked of content based projects. Another platform is medium toward data science and other micro blogging platforms where you could share and showcase your projects. And I've also seen others who've gone beyond to leverage on their web development skills. You've built your own personal blog and you've hosted your projects there. Basically the different many ways and platforms where you can host your projects, the different ways you can come up with innovative and creative projects that you can showcase in your portfolio. You can look at the various role that you're applying for or the industry or in your field or in the domain that you'd want to apply. Again, there are different ways you can come up with ideas on how you can the different projects you can work on. And I believe this being a journey of a thousand steps can help you to have ideas to start working on projects to showcase your content and code based projects. And all these will lead into you getting a job or even realizing more passion towards either particular area and either going into research or exploring other alternative options.

Thank you. I agree like finding projects that are meaningful to you and also alignment.