Accelerating the Digital World Through Machine Learning and AI

with Bassem Asseh,

Head of Sales, Hugging Face

In this episode, Pete is joined by Bassem Asseh, Head of Sales at Hugging Face. Pete and Bassem talk through all aspects of machine learning and AI particularly around the aspects of how the technology is impacting consumers, the new frontier of the digital world, what ethical hurdles need to be overcome with AI, and more.


Key topics in today’s conversation include:

  • The buzz behind machine learning and AI (0:45)
  • Bassem’s background and journey to Hugging Face (5:16)
  • Selling the digital world when it is in its infancy (14:45)
  • Opportunities for businesses in using machine learning (18:41)
  • How the predictive model is impacting consumers (21:15)
  • What does SaaS(ramp) mean to Bassem? (27:41)


The SaaS(ramp) Podcast explores how tech leaders scale from product adoption to enterprise success. Learn more at


Pete Thornton 00:00
All right, welcome back rampants to The SaaS Ramp Podcast. I’m your host podcast Pete welcoming Bassem Asseh, he is head of sales at Hugging Face. Really interesting company can’t wait to unpack this one. Welcome to the show, Bassem.

Bassem Asseh 00:21
Thank you, Pete. Thanks for the invitation.

Pete Thornton 00:25
Yeah, absolutely. I’ve been excited for this one. I know colleagues of mine who I’ve been speaking to are excited as well. Excited to learn more about yourself and learn more about hugging face. And this is a very interesting space that you sit in. So maybe I can get us right into it. Because we’ve seen media frenzy around this particular topic that you’re going to help us learn more about today. And so could you just help me understand in general, what is the media frenzy around machine learning, and AI all about? What have we been seeing in these last three months? And why is it in every feed we ever look at?

Bassem Asseh 01:00
Yeah, this is I mean, this is really an amazing area. I know, it’s, like, it’s usual to say things that are amazing. But this is really impressive. I’ve been in our domain. I mean, in tech for a certain number of probably a couple of decades, maybe more, I will not say how much, but I’ve never seen something like this. Probably every single week, there’s something new that changes things. And probably every single month, there is almost a game changer. If you look at the last year, it was really impressive. Start the year, only people like you and me who are knowledgeable in tech are aware of what AI is. And at the end of the year, even our children, or the teenagers that we have at home, are able to tell you about what AI is, what machine learning is, at least how they can use it, what they feel about it, etc. So it’s becoming something really much more common. If I can say, it’s still something a bit tricky, something magical around it. People don’t know how it works. But it’s really impressive how things are moving fast.

Pete Thornton 02:01
Yeah, that it really is, that’s a great point you bring up about last year coming into the year, you know, like, not completely zero, but very slim. And now it’s in the popular, it’s in the popular culture, you can go to, you can go to as wide of a space as like a technical publication, or you can go to Instagram and see things about it, whether it’s a, you know, whether it’s a mashup of art, or whether somebody’s saying, Give me a poem of admiration for XYZ person. And, and so they’re displaying what they’re able to play with around now, it’s kind of in the common culture right now.

Bassem Asseh 02:38
Exactly. And honestly, like, really, I feel that this is something that is changing things when it comes to the digital world that is being built around us that has been built around us in the last decade or even 15 years. Initially, it was mostly built around software, and software is something where you say I’m inputting this data, and I know what data I can expect as an output. And here with AI and machine learning, it’s a bit more like it’s really different. It’s really different. It’s a game changer when it comes to the digital world that is being built around us. Because it’s going to predict, let’s say 90 95%, hopefully in action 10% of what the answer could be. But the answer is not exactly something that you can expect. That’s what you are expecting from the machine from the models that are out there, they are supposed to predict what you’re asking them to predict they are not giving you a 100% accurate answer. But this is really different from software. Of course, it doesn’t mean that software is fading away, it will always be there. It’s becoming something that is more of a subset of machine learning. But it is impressive how the people who are not in tech are also aware of what it does and the impact it has on their jobs. Like look at teachers, for example, teachers are today afraid not of losing their jobs, but afraid of their job changing because the students will be probably using AI just like the generation before mine started using calculators right? As the hook. So it’s really changing things in the common culture not only in the tech industry,

Pete Thornton 04:09
It really came up again this morning. I have an education background. So I was, you know, a teacher, a science teacher for many years. And so that’s a problem like they can simply ask for, you know, they can give an emotional response, but they can give the actual kind of like, categorical theme, and then it will put together you know, the all the formatting and everything that you’ve asked for. Like it’ll put the essay right back out, just as it was just kind of playing with it like I was interested in playing with it like somebody might have played with software. There feels like a difference though, because if you weren’t to play with software in the 90s this is when I was first touching these things very difficult. Like that’s really hard. I remember I got this little machine and I made a man run across the screen by inputting this. I don’t know what code language it was a toy really but it was the same principles of inputting some form of other language and making an action happen input output like your reference. This is very different. You’re like you’re asking for something and you’re getting something back from, you know, from yet from an artificial intelligence. I don’t know, that was kind of my experience in a broad picture of the last 25 years of feeling software to AI. And you brought that up. Maybe if you wouldn’t mind you had an interesting career over, you know, a longer period of time. Would you walk us a little bit through your experiences and kind of where it’s led? Yeah, sure.

Bassem Asseh 05:33
So, initially, I started my career at Accenture as a consultant. That was back in 1999. So from the very beginning of the internet, I had my very first email, probably 95, when I was in a business school. I started my business school there with a Hotmail inbox. And then 99, started working at Accenture, discovered tech and decided to make the next move four years later, into a software vendor. That was a document management and open source document management software vendor, by the way. And funny enough, back then it was written in Python. And back then no one knew about Python, like, back then everybody was talking about Java, everybody was drinking coffee of love. And back then it was already written in Python, the company was called Nuxeo. It was a French Open Source software vendor around document management and content management. I stayed there for five years, started as a consultant and ended up being the salesperson head of sales. And then I decided to jump and move to another company, which was a British company called Alfresco. And still in Document Management, still open source. So I have to say that I really started my adventure in tech in the open source world and never really worked for a closed source vendor. So really fully open source. I stayed there for almost five years. So these were the 1010 1215 years of my career Accenture first couple of software vendors, five years each, both open source. And then I started working on selling those solutions to enterprise accounts. And something started, like back then we were always thinking about that 10 years ago, everybody was talking about digital transformation, it was the buzzword, everybody was showing very nice slides. I mean, those slide decks were really smart slide decks, every pixel was really great. And those slide decks talking about digital transformation and how the world is changing. And, actually, I ended up wondering about what that digital transformation meant in a very concrete manner, like, how do you build it in a very concrete day to day operationally. And this is where I thought that it would be a good opportunity to go work for GitHub, and simply because of the digital world, men and women writing code, and GitHub was definitely the place where the gold was being built. So I had the impression that being at GitHub would allow me to be at the very core of this nuclear plant, if I can say that the huge worldwide workshop where the digital world was being built simply because men and women were writing code collaborating around code, building, something that was open source, the ability to build an open source software, you are based in the world was to me something really promising. And that’s where I joined GitHub that was almost eight years ago, a bit more than eight years ago, as a salesperson first one of the first three people in Europe, selling GitHub for enterprise accounts back then we didn’t really know if we’re going to going to be sold more to startups, midsize accounts, enterprise accounts, we didn’t really know. But actually, it was really a success, because all developers love GitHub, right? Wherever you are in the world, whatever job you had, as long as you were a software developer, whatever company you were working for, at the end of the day, you were doing something around open source, project, open source communities contributing. And that was happening in GitHub. So we had the software developers loving GitHub, and they wanted to use it not only in the evening, during their own personal, open source projects they wanted they also wanted it during their day job. So we had the community that loved us and the sales team who was able to talk to CTOs, CIOs, architects, leaders in those big accounts, convinced them that it helps companies to grow faster to build software much faster, to go to the market with new applications much faster. So of course, that was something interesting. And during the eight years of my tenure at GitHub, we did that. And, I mean, GitHub announced in August that it had reached 1 billion of revenue, which was I mean, being part of that being part of such an adventure was great. During the last three years at GitHub, I was VP of EMEA. So taking care of the Middle East and Africa. And then in June, July, I started thinking about what’s the next thing that I would like to have, because being in the same company for seven years and a half was, was enough for me, I wanted something a bit smaller, more more adventurous, if I can say, and thought about that thing that made me come to GitHub, which was, what’s the digital world? And where is it? Where is it being built? And what’s the next step in it? And this is where I thought about that idea of software being huge in order to build the digital world. But machine learning was also the next new thing that helped build the digital world. And I’m saying digital words on purpose, actually, because it’s not only about the economy, it’s also about the rest of the society. I mean, whether you are a government agency, local government, and NGO, whether you are a medic, like in a medical school, or whether you are an engineer, any domain you are in the digital world is there in order to offer a certain number of opportunities, certain threats sometimes? And yeah, I was under the impression back then that machine learning was accelerating that digital world. And this is where the opportunity to join a hugging face appeared, then, yeah, of course, I said, yes.

Pete Thornton 11:23
Yeah, this is a very interesting, like route, your experiences have taken you through, because it seems to have been always on to is always very, like progressive technology. And I kind of like what’s on the bleeding edge. And it feels like that’s very intentional on your part. Maybe you lucked into the first open source Python run, you know, because who knows that was so long ago, you might not have known that was going to but it certainly I’m sure it sounded different. Something about it must have resonated with you. Because correct me if I’m wrong, but like if you started at Accenture in 99, and you only spent maybe four years there. So by 2003 2004, you were in a software company that was open source written in Python.

Bassem Asseh 12:06
Yes, that’s exactly around 2005. It wasn’t Python. And I have to say, as a salesperson back then, I was really struggling to sell to customers, a software that was written in Python simply because the customers didn’t have the developers who knew how to read Python, and even less how to write it right. So it was a bit difficult for us to sell it back then. But back then those developers were very smart people. I loved the experience I had with them. They already knew that Python allowed a certain number of things to be done much easier, much faster than the other languages, probably a bit more difficult to get into. But once you are in it, you can do magic, which is less the case with other languages. And by the way, I mean, I didn’t know it back then. So just to be very humble towards those things. I think the engineers have. And scientists sometimes might have some great ideas that are not ready for the world in which they are. And a few years later, the world is ready. And then these ideas become reality. That’s also my impression around machine learning. Because machine learning has existed for not machine learning. But actually, AI has existed forever in papers. And there were some new circumstances that allowed those papers to become reality. And that’s what happened maybe three, four or five, five years ago with machine learning and hugging face was part of this right? Because when hugging started with transformers changed things, they accelerated at least things in machine learning and helped democratize machine learning. And that’s the mission that’s part of the mission that we are, we are giving to ourselves at hugging face.

Pete Thornton 13:46
That is interesting, would you say about your initial experiences with selling Python, because it’s probably going to be the same as hugging face or any of these, like, bleeding edge? What’s Next type companies do the first mover and they always refer to that as the first mover advantage. But in sales, not everybody wants to buy the thing the first time. So like open source Python related, you know, it’s the best, but not everybody. You know, there’s sometimes it’s these other companies that finally take advantage of, you know, BlackBerry’s gone iPhones here, you know, Blackberry kind of had that thing going on first, like, those are interesting challenges. But you’ve seemed to have sought them out, which is, you know, so back to maybe the theme of evolution, when you were talking about digital transformation. There’s nobody, there’s not a listener on here, especially if they’re over the age of I mean, let’s give him credit over the age of 28, who doesn’t know about digital transformation and does not know about this slide that you’re talking about? And if they didn’t hear the Fourth Industrial Revolution and have it beat into them, then maybe they’re not in this. So that is interesting, because you’ve looked at that and it seems to you that we don’t hear about that so much anymore. That’s kind of drifted away. But this next idea, obviously baked around machine learning and AI You’re calling the digital world is that right? And like, because this hasn’t been coined yet, I have not heard a repeated message around this yet. So maybe you could tell us a little bit about what that means to you. Or if you were to coin a word, maybe this is the good public transition to do it. Like, what this space would be called, what they’ll refer to it in three years,

Bassem Asseh 15:21
I would say I was using the word world, even though it’s difficult to pronounce for a Frenchman like me, but the word world is simply I’m using it because it’s way to say that what we are talking about is not only about enterprises, or startups or the economy, generally speaking, or business cases, business situations, it’s more about everything that we offer, everything that we are using on a day to day basis, I mean, you drive your car, you are using something that is built on using software and using machine learning use your phone, of course it is you use your you open the fridge at home to grab something to drink, that there might be at some stage something more than just the software in it. So this has changed. And of course, this also happens for NGOs. It also happens for governments, agents, and government agencies. So it covers all these, all these interactions that we may have on a day to day basis. The interesting thing with with machine learning is that if I had to compare, for example, my experience at GitHub and my experience at hugging face at GitHub, when we start when I started eight, nine years ago, even 10 years ago, you were leveraging a huge feature that is called pull requests that every developer today loves. But imagine software development without the pull request, because there was a world where the pull request didn’t exist. And that world existed. And GitHub made the I mean, we had solutions like SVN or ClearCase, people who are beyond 35 probably know about those things. But GitHub came and replaced these things. And when I was there trying to sell GitHub and sell the pull request to enterprise accounts, clearly they had their SVN, they had their clear case, and they were happy with it, they had it for 10 years, 15 years sometimes, and they didn’t really want to change. And potentially they could still keep those old solutions for another 234 years. And then they ended up adopting the pull request, which means adopting GitHub, that was the software development journey. If you look at the machine learning journey that is just starting, I mean, it started three, maybe three, four years ago for many people. And again, last year for another chunk of people, those solutions, and those technologies are not simply replacing an existing thing, they are opening brand new landscapes that didn’t even exist before. And this is, I mean, this is really so exciting to be part of. And I’m really humbled in front of that, because, like, I’m not sure that everybody knows what it’s gonna give in three years. We feel that we are all together, we feel that it’s going to give something really great to reach certain frontiers and technological frontiers much faster. With certain threats. I have to admit it. And there’s sort of like an ethical dimension when it comes to machine learning that needs to be taken into account. But there are also so many opportunities. And yeah, I mean, I’m always thinking about the students that didn’t have those that were not authorized to use calculators back then when calculators started. And now they are, I mean, now it’s business as usual. Right? It’s teaching as well, you use the calculator? That’s it? Yeah, yeah. That’s mostly it. I mean, the opportunities you need to take care of the threats that are coming with every opportunity. But the opportunity is huge.

Pete Thornton 18:45
To make this like, like kind of a practical takeaway for anybody really kind of unpacking this topic for the first time, is there any one or very interesting, but specific opportunities using machine learning or AI that you’ve seen or that you’ve thought of, or just being in this space all day, every day has kind of come to your attention?

Bassem Asseh 19:03
Actually, the idea is that because in this digital world in which we live, we are producing data, and the people who are using those data, they can use the data in order to predict what’s the next thing that we might need. For example, if you are listening on Spotify, or Deezer, or the equivalent of Apple, Apple music or whatever, the solution you are using to listen to music, because you’ve listened to this and this, then most likely you will like this thing. So the recommendation model that is behind that feature pushes you towards listening to that new song. This is a model that is a machine learning model that uses the data collected from all of the users and predicts what you might like. Of course, you might not like it, it’s okay. That’s the rule. That’s the name of the game. It’s only a prediction. It’s not 100% Sure, this is all So the difference between traditional software and machine learning. So everything related to this kind of recommendation is a great, very practical, very practical use case of machine learning. And there’s a certain number of other things, including, for example, if you have, if you are collecting data in a power plant, it probably also allows that data will probably also allow you to predict what kind of failures you could have in your power plant and potentially fix the issue before the issue happens. So you would reduce the failure time. And, and this is also great from an industrial point of view, for example, I’m giving two examples. One very, very, like something that we use on a day to day basis thinking about Spotify, and the other is a bit less what we use on a day to day basis, even though it’s something very important. If we want to have energy, it’s better to avoid the failure in the power plant. So these are a few things, the idea is that the data that you collect, are, are being sent to a model and the model with the compute power that is behind it, is able to predict what can be what can be happening afterwards. That’s more clear to know, as it’s not very technical, I’m sure that the people who are much more technical than me will be providing examples that are a bit more complex to understand or to use, but that’s mostly it.

Pete Thornton 21:18
No, it paints a picture because that’s like a consumer driven example, something that the minute everybody looks right back down on their phone, they’re like, oh, yeah, that’s right, they did recommend that you’d like this that, because these things are telling us like these things, I’ll have a thought in the day. And by the night, I will not have spoken out loud or texted about it, I promise. And by that night, this showed me something that I wanted to say. And sometimes I get a little weirded out by it. And then sometimes I’m like, love that I didn’t even have to even like to come to, you know, it just kind of understands where you seem to be at or, you know, that’s a very interesting consumer example. And on the other side, you know, if you’ve seen the Netflix special recently, Chernobyl or something like that, you’re like, Yeah, bring it on, let’s get some safety, let’s get some predictability. Let’s like, you know, around these industrial use cases. So those were solid examples.

Bassem Asseh 22:05
So yeah, this is what I mean, if you take your phone in the morning, and you search, you open the Search field, in order to find the application, it will offer you the opportunity to go directly to the application that you mostly use. In the morning at that time, when you start using your fork, that’s also just the usage of your own data. It facilitates the usage of some of these things that we are using all day long, whether it’s a car, or whether it’s a phone, but it’s yeah, it helps predict things and simplifies your life. And, but also, I would insist on the fact that we need things to be done in an ethical manner. And this is where open source helps. I’ve always been in the open source area and the open source part of the digital world. So I will be evangelizing about open source, of course. But yeah, having machine learning being done in an open source flavor is also a way to make sure that we are all able to understand what’s happening in that machine learning black box that is helping me simplify my life on a day to day basis. So open source is, in my opinion, critical in the way we are building machine learning today and tomorrow.

Pete Thornton 23:18
Yeah, absolutely. We stopped your journey. We stopped your journey when you arrived with a hugging face. So where does hugging face fit into this digital world? Like what? Oh, yeah, what things are, have been propagated there.

Bassem Asseh 23:30
So I think the probably the most interesting thing about tagging faces, the community that it’s been able to build, and the way it’s able to democratize machine learning simply because if machine learning is built around models, datasets, and the compute capacity for those models to learn from the data sets and compute capacity to predict, to predict what is expected in an application, that’s exactly what tagging phase is providing. So it’s collecting all of those models that are open source made available there. It’s collecting all the data sets that are open source available out there. And it’s giving the opportunity for machine learning engineers and data scientists to access all these things, models, datasets, and some features that allow models to be trained, and even features that allow machine learning engineers and data scientists to demo what those models are able to provide in terms of outputs, and in a very simple manner. So these are without going into detail to all the features but that’s mostly as the platform that allows machine learning engineers and data scientists to access the models to modify models to build models, to share models with the rest of the community. So it’s mostly about building the community and enabling it to take machine learning to the next step.

Pete Thornton 24:50
So this is interesting because it’s open source and because you’re building community by hugging face, this is where like enthusiasts essentially can come whether it’s based on what they’re doing in their day to day work or what they were doing in the evening. Because this has a familiar like pattern to it from before, because these are what people were doing with GitHub when they were more concerned with building software. And now if they’re more concerned or interested in and contributing to the, you know, machine learning AI space, then they would come to hug their face, be part of that community, and even be able to kind of demonstrate what they’re doing and learn from others more important. Exactly.

Bassem Asseh 25:25
That’s exactly it sharing and learning from others, building on what others did, and giving back to the community. That’s the thing about open source. And of course, it allows it, so it allows projects to go faster, but also for potential difficulties on some projects to be fixed faster, because it’s done openly. So people can come into a certain project and provide some help. And also it allows things to be like, having these things done openly means that you are able to check if there are biases that need to be fixed. Because if we are talking about predicting things, you predict things based on data, but data can come with biases. So these biases also need to be way up, we need to be able to be aware of those biases and fix them. So that’s why open source is that important on both sides, both on the sharing and moving fast, but also on the fact that you are able to identify biases and help fix them.

Pete Thornton 26:27
Okay. Yeah, that makes complete sense. And it’s interesting, because I kind of understand the model. And then I understand kind of like how that community can build. It’s a long tail sometimes like, because it’s, it’s something that’s but this is where all of his really great things come from, it used to have to be physically present for these things that was like, you know, you had to be in the Silicon Valley bubble, to kind of understand what was happening in that garage. And now that we can do this digitally and remotely, it can expand and now there’s a place for people to come who have this kind of enthusiasm for this up and coming. Massive space. So it’s super fun to share your enthusiasm for it. Like, it’s really interesting. And you haven’t doled everything out or whoever’s in charge of it. Just tell them congratulations, anybody listening, go check out hugging faces and just hugging faces. As soon as you see it, you’ll be like, Okay, it’s just it’s a fun little look and feel and brand to it. So people will either recognize the icon or know what you’re doing. These are people I speak to so maybe it’s a subset of the population, but know what you’re doing and be very excited for the space in general. So anyway, congrats on the success today.

Bassem Asseh 27:38
Thank you for the best words, I’ll tell them. Hey, yeah,

Pete Thornton 27:41
there’s gonna be one person back there, but I knew they’d love it. Okay, so this is a SaaS ramp podcast. And this would be an interesting question, because, you know, it’s, we’re elevating a little bit or bending in the SaaS realm. But the question is, what does SaaS ramp mean to you? So I’m really intrigued by your answer on this one.

Bassem Asseh 28:02
Sure. So actually, I was thinking about it what? So we are talking about software as a service, right. And I was wondering, what would be the word or acronym that we will use when machine learning will be much more democratized? And we will be all using it, including software developers by becoming like machine learning as a service? Or, which means that you would probably need to adjust the name of the podcast? Maybe not right now. But Oh, right. And so it’s interesting to keep on your radar. Yeah, I was. Yeah, to me. I mean, we have been all in the world before SaaS even existed. So you needed to download the software, put it on a server, it was your own server, because there was nothing like nothing that we could call cloud today. So we moved from that world of software that you needed to download and install on a server to a world where you are using software without needing neither a software, nor a server, nor anything to download. So you’re just using SAS and probably tomorrow, or in the next few years, we will also be using machine learning without needing any character like any ability to know the details of how to download and install the software on a server. So probably the ramp in the future for SaaS is also from trying to transition from SaaS and itself to machine learning more and more software. Software, of course, will always be there. But my feeling we are a certain number of people thinking about that will be a kind of subset of of machine learning. And that’s the next ramp I think for software as a service.

Pete Thornton 29:42
Yeah, that’s it, so it’s so interesting. So it’ll be welcome back to the molests medulla last rap podcast. Yeah, and that will be an exciting time if, when it is commonly agreed that we’ve moved into that realm now we’ll have zero problem Moving from Episode 52 When it was told to episode, you know, like 200 when we actually wrote the transition. Bassam, that’s a wonderful insight. I really appreciate it. Thank you on behalf of the audience because a lot of us have just been curious and wondering even within the space and getting a little bit more blips this whole light consumer movement over the past year has you can’t you know, you’re living under a rock if you haven’t seen it. So being able to have you on and understand your experiences and see the evolution is quite eye opening. Really appreciate it.

Bassem Asseh 30:30
Thanks a lot for the invitation. It was great having this conversation and thank you again.

Pete Thornton 30:34
Brilliant. Have you again on soon.