Jake Aaron Villarreal: I'm Jake Aaron Villarreal, born and raised in Silicon Valley and here to take you behind the scenes to share what it's like to be a startup founder, the journey they're on, the problems they face, the products they build that are transforming industries. I'm excited to have with us today Dr. Amr Awadallah, co-founder and CEO of Vectara, a company that is enabling organizations to leverage trusted Gen AI in business applications by reducing risks from hallucinations, biased, copyright infringement, and intellectual property pollution. Um, welcome to the show.
Amr Awadallah: It's my pleasure to be here, Jake.
Jake Aaron Villarreal: Great. A little bit more about Amr. He previously served as VP of developer relations for Google Cloud prior to joining Google in 2019. He also co-founded Cloudera in 2018 and was CTO for se for for 11 years working closely with enterprises around the world on how to ingest and extract value from big data. He famously coined the concept of schema on read verse schema on write. He also served as vice president of product intelligence engineering at Yahoo from 2000 to 2008. In fact, his first company was acquired by Yahoo which was a search engine for online product information. He received his PhD in electrical engineering from Stanford University and his bachelor and master's degrees from Cairo University in Egypt. Well, we're excited to talk with you today here and you have a lot in your background that's I'm sure going to be interesting for a lot of the listeners. Uh before we dive in here, where are you calling in from today?
Amr Awadallah: Today I'm actually calling from Carmel Highlands in California.
Jake Aaron Villarreal: Hundreds of AI startups are launching every month, battling to build their founding teams. As a leader, your job is to get results. When it comes to hiring, that's where it gets tough. So, you go out and you try a recruitment firm, but they don't understand your story. They're off target. And when they send you candidates, it's a waste of time. We believe you should never have your time wasted. That's why we launched Match Relevant because your story is more than just an open role. It's your founder's journey, the problem you're solving, the product you're building, and why it matters. When we work with companies, we make sure we understand your whole story. So, we go out and do a search. We're on target. It's worth their time. They're interested. And more importantly, it's worth yours. And when it comes to hiring engineers, we work to make sure we get it right by deploying a team of seasoned CTOs that have built some of Silicon Valley's best companies. They can collaborate with you in the technical interviewing process. They can be a sounding board or they can run it for you. When it comes to building teams, there's no time to waste. Let's make it count. If you have a role that needs to be filled, book a time with a hiring guide at matchrelevant.com and learn how we do it.
Oh, nice. I like that area.
Amr Awadallah: Yeah. Uh, I mean, I typically live in Palo Alto, but we have a heat wave right now in the in the Bay Area, so I'm escaping escaping the heat as as much as I can.
Jake Aaron Villarreal: Good strategy. I like that. Yeah, we're located down in uh Laguna Beach and luckily we have a little bit of this offshore winds and things are a little bit cooler down here, but...
Amr Awadallah: Exactly.
Jake Aaron Villarreal: Uh, good. Well, thanks for joining. Um before we talk about you and your company, what you're building today, give us a little background on kind of your origin story kind of where did this start and how did you get into technology?
Amr Awadallah: Yeah, so I would say it's really curiosity like uh the origin is all comes from curiosity. I think science and engineering is always driven by that uh trying to figure out how things around us work and then how to make them work better. So when I I I'm from Egypt originally. Uh when I was very young, uh like six years old, seven years old, I liked to just break everything that my dad built. So he would buy a new fridge, a new TV, a new washing machine. I would open it up and then reassemble it. And sometimes I am successfully able to bring it back to life and sometimes it just breaks. I I drove my dad nuts doing that. But I think that's the origin. The origin comes from very naturally from a young age. I always had this curiosity for how things work.
Jake Aaron Villarreal: Yeah.
Amr Awadallah: Yeah.
Jake Aaron Villarreal: That's great. Well, the curiosity led to you becoming an entrepreneur and founding companies. Um, did you have that sort of makeup growing up? I know the problem solving and taking things apart technically makes a lot of sense, but talk about the entrepreneurial side. Were you the type that were selling products or services at a young age? Like where did that part come from?
Amr Awadallah: Yeah. So, believe it or not, it wasn't part of my DNA at all. Like when I was young in Egypt, uh it's very hard to be become an entrepreneur in Egypt. You need to know somebody who knows somebody. We don't like the the law system is not as fair or nice as it here in the US. Uh connections uh matter more than aptitude. So, I wasn't really dreaming of becoming an entrepreneur back in Egypt. My dream was just to become a professor like my my dad. My dad was a professor at Cairo University. Uh and he would tell me since I was very young, okay, you're going to be an engineer when you grow up, but you're going to be a professor engineer and you're going to teach engineering at the at the at the university. And I did indeed get my uh bachelor's and master's degrees from Cairo University and then I came to the US in 1995 to get my PhD from Stanford.
And then uh Stanford was a very pivotal point. So Stanford shifted me from being somebody who wants to become a professor and seek a career in academia to being somebody who wants to become an entrepreneur and that comes from a number of reasons. The first one is just landing. When you land in the Bay Area, as you know, Jake, having been here yourself, you see all these companies we hear of. You see Nvidia, you see Intel, you see Google, you see Oracle, you see LinkedIn, like they're all right there. Twitter, Facebook, you name it, right? So, right away, you start grounding yourself that these companies we heard about, they're actually real. They're there. So, that's number one.
Number two, at Stanford, they uh they uh from day one, regardless of your discipline, if you're in finance or law or health, they teach you about entrepreneurship right away. How to make a business plan, how to make a pitch, how to approach investors, how to do sales and marketing, branding, messaging. Uh so that equips you with the raw skills that makes it more tractable for you. It's like like I I used to think starting a company is very hard and I was scared of starting a company. It is hard by the way. I don't want to admit how hard it is to start the company, but part of that fear came from my ignorance of the tools and the techniques you need to make something like that happen. So, they give that to you in a silver plate. So, that's number two.
And number three, Stanford keeps inviting all of the entrepreneurs from the industry like the the founder of Google, the founder of Yahoo, the founder of HP, the founder of Oracle, the founder of Nvidia to come in and talk at Stanford. So you get to see these people and when you see them in real life, not just in a YouTube video and talk to them in real life, many of them are very and that's one of the key I think attractive things about the Bay Area. Despite of their success, they're still very humble. They're very humble and they're very uh confident at the same time. And that makes it for you more achievable. Like I can be like that. I don't like these guys don't have wings. They don't have like halos around their head that make them 100 times stronger or smarter than any of us. They're just like they have perseverance, they have grit, they have determination to get things done and they have great ideas. And so I shifted like Stanford really I usually joke and say Stanford corrupted me from the teaching mission to become a professor to instead become an entrepreneur. So I give all the credit to Stanford actually.
Jake Aaron Villarreal: Yeah. You just never know where your education is going to take you and the influence it has on you when you go.
Amr Awadallah: Exactly. That's why we go to that that and by the way that's one of the things I love about the university system here in the US is where they let you study many many things like in in in Egypt at least and I don't and most other countries you get to specialize you get to pick one thing and like I'm going to go to the faculty of engineering and that's the only thing I'm getting no in the first couple of years in most universities here in the US you get to explore right you see economics and you see accounting and you see engineering and you see um law and and that makes you find your true calling right it makes it a lot more better for you to find your true calling than having your mom and dad tell you when you grow up you're going to be this or you're going to be that.
Jake Aaron Villarreal: Yeah, 100%. Well, you had some early successes in your startup experience. Uh, Epivia being acquired by Yahoo and um if I recall correctly, that was kind of the the beginning of where search was happening as well as e-commerce and being able to take products and get more visibility to where they were then where do you go to buy it? So yes, you were really on the forefront of technology early on. The company that I think is really interesting because we just read so much about it was Cloudera and it was almost like the company of Silicon Valley for a moment where everyone knew it. Everyone thought it was an incredible company. You guys went public. You really scaled up. Uh, and for the listeners that that's a dream for them, can you just before we talk about your current company, Vectara, and get into that, can you share a little bit about what you thought the experience was going to be like when you started and really built and scaled and went public to what it really felt like in that in that experience.
Amr Awadallah: Yeah. So, first I will I will note is I like to remind all of us starting companies is hard. It's not easy and the chances of success is low actually. So, you need to be brave and you need to be uh willing to take that risk. Willing that I'm willing to risk the stability of a big job in a big company, go start my own company when I'm paying a fraction when I'm being paid a fraction of what I'm being paid at that big company. But now I have a big chance of number one delivering on the mission, the vision, the solution that I am believer in and then number two uh having the success that that might come with that as the company succeeds.
Definitely taking a company from four people. Cloudera was four people at the beginning. There was four co-founders. There's just the four of us. That was in um in summer of 2008. We got together. We started the company in October of 2008. Officially the first round of funding came then in October though we started a few uh weeks before that. And yeah, that company went from four people all the way to thousands of people and going public on the New York Stock Exchange. And I cannot uh stress how rewarding it is actually to build something like that. Uh on the day that we were in the New York Stock Exchange, they have this very old tradition to give you a hammer and a bell and you have to ring the bell with the hammer. Like you just hammer the bell and that launches your stock now for public trading. It's a very old tradition because today you don't need ring bells and hammers. Today it's all done by computers anyway. In the old days, they would ring that bell and then people start shouting out the prices and and so on, which is funny that they kept that tradition. But when you when you when you hit that bell, launching that ticker symbol, I got shivers like all of my hairs in my hands were were standing up and I I can feel it throughout my whole body. It's like, wow, this company now is born. This company now is is out there. It's independent. It has all governance structure, shareholders, and it might survive and live longer than I would live. And uh I frequently would tell people that the feeling you get from that is very similar to your first kid or f first girl or boy that you have. It's like you see this thing that came out like wow I helped make that and same level of thing which is very very rewarding but I don't want to diminish uh the importance of you being ready that it's hard there's going to be ups and downs there's going to be tears there's going to be joy as well but it's not easy you have to keep in mind that it will require a lot of grit a lot of perseverance a lot of determination to make it succeed.
Jake Aaron Villarreal: Yeah, and a lot of people to make it succeed too and the right people about the team yeah like...
Amr Awadallah: Some entrepreneurs think it's about the idea. No, the idea is only 1% of the of the of the story. It's how you going to execute on that idea. And how you execute on that idea is going to depend on the amazing team that you build around you that's going to make that happen. And if you're from day one, um I don't want to say like that doesn't work because there's people that do that like Elon Musk. It's all about him. We we know that and he still succeeds. But I'm a big believer that it's really all about the team. And I'm I'm a very big believer that a team a great team of people will always beat a team of greats. So what what I mean by that is a team of greats is a bunch of geniuses that don't know how to work work with each other, right? So they they're all very smart. They think they are God's gift to creation. They think they're the smartest people on planet earth and they have this uh uh uh arrogance in them. So they don't have the humbleness uh and hence they don't work very well with each other. I think a team a great team of good people that are very confident and working together and really optimizing for the team will uh always win against a team of greats and that's the kind of cultures I believe in and like to build around me. So I agree with you 100% Jake that it's not about me it's about the team I'm able to assemble around me to make it succeed.
Jake Aaron Villarreal: Yeah. And you don't have to be the founders to actually have that experience either or be there to see the rewards. Yeah. If you're there in the beginning, that's why a lot of people join startups where they might not have the idea, but they can be part of that team and and get some of the fruits of the labor in the process which then propel them to be their own founder as well.
Amr Awadallah: And in fact, we did invite a lot of the founding team members, the like the 10 the first 10 20 employees that joined us at Cloudera. We did invite them to come to the IPO and witness this special moments because without them that would not have been without everybody would not have been possible. But specifically your first core group of like 20 30 people they become the genesis they become the seed of that company in terms of culture in terms of uh work ethic in terms of execution in terms of focus they are the the seed that make that happen so you you really want to reward them very well so that they be because they were loyal to you and you need to show them loyalty back 100%.
Jake Aaron Villarreal: Well that's great. Well, what a story in itself and just the the accomplishment to go through that, the experience and the confidence you can bring with you as you go forward um can lead to many other things. We're going to talk about Vectara today, your current company.
Amr Awadallah: Actually, before we switch to Vectara, I apologize to interrupt you. Just I forgot to mention about Cloudera is how did the idea for Cloudera come around actually. So, Cloudera was about big data like how can we analyze big data in a more flexible scalable way. So previous systems were databases. Databases were very rigid. They're very structured, very rigid. Uh but they're hard to evolve with the different types of data that you have. And databases have a language called SQL which you use to ask questions but it's not always the perfect language. Like sometimes you want to ask questions with other languages and that was the genesis for for Cloudera. Now how did that idea come to me? That come that came directly from my work at Yahoo. So while I was working at Yahoo I was towards the later years I was responsible for business intelligence uh meaning reporting back to the business how they're doing and how can they optimize the products better to increase retentions of users and engagement with users and as part of that we had to build a very very scalable flexible uh data processing system that can do that and none of the legacy data uh bases were were were hitting the spots they they were not flexible enough they were not scalable enough and they were very very expensive to be honest uh and that's where the genesis from for Cloudera came to be. So one of the best the summary of that why I say that is one of the best lessons for uh anybody listening to the show right now in terms of what is the next big idea I should be working on usually comes from your job it comes from a problem that you're facing yourself and because of your experience and knowledge of that problem you come up with a very unique and differentiated way for how to solve that problem way better than others. Uh so always be looking at that if you're thinking about starting a company. Always be looking at your job for clues for what's the right problems that you'll be going to solve.
Jake Aaron Villarreal: Yeah, it's great. Thanks for adding that to uh to the conversation here. If you look into business, yeah, sometimes it's about understanding the problems you're trying to deal with in your own role and that becomes the idea that ends up being your next startup. But it's really about understanding what problems are in the market that you can help solve. we talk about your current company of Vectara. What what is the problem you're looking to solve today?
Amr Awadallah: Yeah, so there's two problems. There's the meta problem and then there's the technical problem. So the meta problem is uh AI is finally here. AI is finally here. Meaning technologies that can uh reason, read, plan, think, generate in the same way that we do as humans. that and and and our goal is to help companies leverage that capability within their organizations. So so our primary goal at a very high level is what we refer to as trusted generative AI. How can I use my generative AI within my company while maintaining the trust with that technology to help me do the right things.
Uh the more technical explanation for what we do uh for the listeners that know the solution is called retrieval augmented generation or rag as a service and retrieval augmented generation is this technique that allows geni to produce responses and take actions that are grounded in your own data right so chat gvt for example is what was trained on the web data what's available on the web Wikipedia and New York Times and number of other uh sources of contents uh we when we're doing this for a customer it's based on their own data that we're doing that and then helping them find very good actions solutions responses as a function of that data by leveraging uh geni within rag when you're doing this technique that's called rag meaning geni for business there's three key problems that the businesses really care about that you have to solve if you don't see solve these problems within geni then they're not going to be able to deploy you they're not going to be able to use you the number one problem is accuracy of the responses uh which is this hallucination is issue uh the large language models which are these techniques behind unfortunately uh they have this very small probability of about 2% where they might make up stuff they might make up completely made up facts that are not true but say them in a way that appears to be so correct to the to the extent that it tricks us as humans we think that is the right answer and there's many scary stories last year about lawyers ers that lost their jobs because they believed the AI and just used the AI to build their their their lawsuit or um airlines that had to send sell a ticket for $1 because uh a user tricked the large language model into offering that or somebody getting a car for free by tricking a car dealer uh large language model Jenny I bought to do that. So the that's all a manifestation of this hallucination issue. So that's number one.
Number two is security. These systems need to be secure. uh you you have to make sure that they don't reveal secrets that they should not reveal or if they're revealing them, they're only revealing them to the right person with the permission level to be able to see these secrets. So, believe it or not, that's the number one concern actually that businesses have in terms of using Genai within their organizations.
And then the number third number three problem is um uh explanability. If if you're using GI if you're using AI in general actually not just generative AI and using it within a regulated industry business like finance insurance uh legal accounting health manufacturing telecommunications government all of these regulated businesses uh most of them require you to explain the response. You cannot just tell me here is the answer. You cannot just tell me here is the plan. You cannot just tell me here is the action plan that you need to execute right now. You need to tell me why. explain to me why how you came up with this plan. Why is this the right plan? Which of my documents and PowerPoint presentations and and uh PDFs did you depend on to come up with these conclusions? Uh so that's a very very key problem as well. So we solve all these problems. We solve the the problem of hallucination and accuracy. We solve the problem of security and permission access control and the problem of explanability and explaining the responses that you get back. Hopefully that makes it very concrete to you what we do.
Jake Aaron Villarreal: Absolutely it does. And for the customer out there hearing that, the benefits that they get by using your company is going to be what...
Amr Awadallah: these these three things I just described right now. So they can try and go and build that with off-the-shelf technologies uh build these geni rag pipelines using open source or whatever but they're going to find it's very easy to prototype like they can build a prototype uh a sample application very quickly but once they launch that sample application to their users whether these users are internal user or external users they will start to complain with them to them right away about these things they're going to start to complain wow these are inaccurate responses why you giving me this madeup answer that is not correct contracts and it just takes one madeup answer to lose trust completely in the system. Uh so you need to be very very careful not to produce madeup answers.
Number two, they're going to start to complain about the security. Oh my god, this employee was able to ask this question in this way and they got back data that only the CEO or the CFO is allowed to see. Oh my god, they got back the salaries that only the HR manager is allowed to see. Right? So you need to solve that problem. They will you hear them complain about that. And the third one they're going to complain about is this system is not explaining to us how it came up with these conclusions. We're very worried about that.
There's a couple of other secondary problems uh that we also address. One of them is copyright infringement. Uh so sometimes these large language models can produce paragraphs that are verbatim the copyright of somebody else. Uh open uh New York Times for example right now is suing OpenAI uh because of that because sometimes Chad GB produces paragraphs that are literally paragraphs from New York Times articles as as is. So one of the things we do in our system is we suppress copyright material from being uh generated. Another issue is the issues of bias or toxic toxicity where the system might respond in a way that's very biased against you depending on your race or religion or whatever or gives back a response that's very toxic in nature. It answers in a way that is very makes you uh feel like it's a rude uh AI assistant talking to you. So you want to address these things as much as you can.
So this is just a flavor of some of the things that we solved. Now I want to give you a concrete example that helps you uh see how our customers leverage our technologies. This one is from a company called Sonosim. And Sonosim is building something amazing. Actually what they're doing is they're helping uh radiologists use ultrasound machines better. When you're using an ultrasound machine, it's very hard to configure that ultrasound machine depending on the user. Right? If you're scanning for pregnancy, a heart, a muscle, the lungs, male, female, age group, race, you have to calibrate the machine differently every time and configure it differently every time and aim the scanning device differently every time depending on what you're doing. So only the expert radiologists know this. The average radiologist uh they struggle with this all the time. So what this company has done uh Sonosim, they collected uh the ground truth. They collected all of the manuals of the ultrasound machines out there, all of the best practices for how to configure an ultrasound machine depending on different uh criteria of the patients, uploaded all of that data into our system as is all of these documents into our system. And what they get back is an AI assistant that you can ask now very deep questions about um ultrasound machines and you can tell it, hey, I'm trying now to do the scan for this type of patient. This is the type of symptom I'm looking for. What should I do? sorry and this is the type of machine I'm using the model and the manufacturer of the ultrasound machine what should I do and it'll tell you back this is the configuration that you should put and it gives you the perfect configuration so this makes the average radiologist become an expert radiologist just like that this same pattern will repeat for every single discipline you can think of every single discipline we have today uh including doing a podcast with me right now or hiring uh trying to do recruiting or hiring will have AI assistants that help the employees become a lot more productive in terms of how they do these things and eventually some of these AI assistants as they become more accurate and more dependent they will start to evolve to what's called AI agents instead AI agents will automate some of the tasks right they will they will take care of some of them like for example how can I config the ultrasound machine would say here's the configuration would you like me to just apply the configuration for you and it will just apply the configuration directly to the machine for you or if you're trying to hire a candidate it would see what you're saying and say okay I figured out what you're trying to You're trying to find a machine learning engineer that knows PyTorch that lives in this region, blah blah blah. Here are four candidates that match already what you're looking for. So, you're going to see that pattern repeat over and over and over again across all disciplines. And that's exactly what the Vectara Rag system empowers.
Jake Aaron Villarreal: Wow, that's amazing. Well, there's going to be a lot of opportunities for you to talk about your product to customers for sure. Um, but when you're out there selling, when you're out there reaching your market, who cares most about your product, who within the organization really wants to hear your message?
Amr Awadallah: So, two personas. Two personas. There's the persona which is the business buyer meaning the head of the law office uh the head of the customer support department uh the head of the products for the ultrasound machine like Sonosim that we talked about earlier that want to make their products want to make their customer support want to make their legal workflow a lot more efficient a lot more productive and they want to leverage because the board of directors keep telling them you have to use geni this way or that way they want to leverage to do exactly that to become a lot more efficient and get a lot more out of their employees. So that's one persona that you talk to and then you tell them, hey, we are the only ones that solve these key compliance issues around uh explanability, around accuracy, around security so that you're able to deploy such a solution within your organization. So that's number one persona and message that we deliver.
The number two persona is the developer. the developer that's going to build this capability into the existing workflow, the existing the existing law office tools, the existing customer support platform, the existing uh product as in the case of Sonosim that is being sold to their end customers. that developer also needs to get comfort that our solution is uh going to deliver on all of these promises of accuracy, security, explanability but also can do it in a way that is easy for them to consume. And in fact that was one of the key lessons I learned from my Clouda journey at Cloudera. I built a very powerful product. We built a very powerful product called the Cloudera data platform or CDP for short. But the problem is it was very complex. It was a very complex product. Very powerful, very complex. I called it like a Lego blocks, like a box of Legos that allows you to take these Legos and you can build whatever you would like with them. But you have to know how to put the Legos together. If you don't know how to put the Legos together, you're not going to be able to build whatever you you like with them. Uh, and very few developers actually they have the aptitude to do that. Most developers, they simply want to use a ready-made solution to just make their job work. So for example, an equivalent company to us that did better than Clouder actually is a company called Snowflake. and Snowflake they came in let's solve these problems with data uh being more agile more scalable more flexible but let's solve it using SQL and extend SQL in a way that makes it easier for the average developer to do it and and lo and behold actually Snowflake built a much bigger company because there is way more developers in the world that prefer readymade solutions they can plug in and just solve the business problem they're after versus having to go and build reinvent the wheel and build that solution from scratch and uh that's one of the key messages that we are focused focused on and and not not just messages but key tenants at Vectara that we're focused on is building products that is so easy to use by the average developer. You don't know need to know AI and machine learning and natural language processing. No, it's a very simple API. You put your data, you issue your prompts or questions and everything else just works.
Jake Aaron Villarreal: Wow. That's amazing. you know, companies um when you talk about the ones that really do well, it's about innovation and marketing essential to making any company successful.
Amr Awadallah: Yes. And ease of use, I would add absolutely ease of... people forget ease of use. Sometimes you might think as ease of use as part of innovation, it's not necessarily like again Cloudera was very innovative. We had very very innovative. My previous company was extremely innovative, but it wasn't easy to use. It wasn't easy to use. And and and that was like again I'm saying that's a very key lesson I want to give to all of you listener right now. Always start with ease of use first. Like think about ease of use and then move backward into how can I get the innovation to fit within that not the other way around because if you don't have ease of use as a core tenant from the beginning it's very hard to add it later on.
Jake Aaron Villarreal: Yeah. Makes a lot of sense and that's one that every company we talk to I think tries to strive for. It's maybe one of the harder things to do though when you're talking about taking the technology and making it easy to access, use, or just in general integrate. Not as easy as you would think or as everyone would do it, but I really that's that's a great note. Yeah. Um when you go out to market with a new product like you have, maybe not new today, but it's been out for a little while now. Um what's worked for you? like you know you've got email marketing, you've got outreach with enterprise customers, you got your own founder network, but for startups that don't have a lot of that like what are some strategies or some areas that you can be helpful in there?
Amr Awadallah: Yeah. So my answer to that is all of the above. I mean you should be using every single resourceful resourceful tool at your uh disposal to achieve that and it's that is especially when you're starting from zero that's one of the hardest things to do even for me like as a repeat serial entrepreneur it's still very hard to do that and uh you need it depends on your situation of course if you're doing consumer versus enterprise software and even within enterprise software it depends if you're doing it proprietary way or the open source way this and if you're doing application layer uh enterprise software like Zoom or like notion or are you doing platform layer enterprise software like Vectara or Cloudera or Snowflake or data bricks like all of these modalities change the way that you do your your messaging and marketing but the end of the day it boils down to uh are you solving a real problem that your customers care about. That's number one.
Once you figure that then who are these customers and by the way product market fit sometimes people think the product market fit is about building the right product for the customers. In many in many cases actually product market fit is about finding the right customers for your product like you came up with a very new innovative product that solves a problem in a very unique way but not everybody is going to see that. there's some some uh uh customers that will see that way better and product market fit is about finding them actually uh and the effort goes about finding and of course sometimes your product could be a bad product and and as well like that I don't want to diminish that but the hard part is finding the right customers the right beach head the right group of people that will give you the oxygen will give you the feedback will help you build that product that will become a a a bigger massive scale product after you cross the chasm and the tornado as as you the gorilla marketing curve thing which I'm sure everybody is is is familiar with.
So how do we achieve that? We achieve that by using every single tool at our disposal. Like I reach out to my network. Uh we do keyword advertising in Google. We do social media uh marketing. We do social media just engagement. Uh for developers, we do things on gut uh sorry GitHub. We do things on hugging face. We do things on Reddit. We do things on hacker news. So yeah, you need to speak in all of the places where you can get the message across and leverage every single tool at your disposal. Last year we we're very we were very focused more on because we didn't really have a lot of funding. We're very focused more on very frugal, efficient ways to get to them. This year we raised a bit more funding. So we're going to also start doing um um attending conferences and speaking at conferences. That can be expensive if you're not invited. you have to pay like 15k 20k to get sponsorship and then get a speaking slot or do a dinner for CIOS and CTOs in different cities around the around the US. These things are a bit more expensive. You don't want to do these early on in your career in your startup's career. You want to do it as the startup proves its reputability. Then you want to start doing these because they are more expensive. Uh but that's that's another way that you can reach your customers. So the summary answer your question, Jake, and I'm sorry I'm not giving you a perfect answer here is all of the above. You need to do everything you can do to get those first 100 customers.
Jake Aaron Villarreal: Yeah. Well, what I'm hearing is it's a lot of effort and work, but you got to do it all. You can't expect one silver bullet to kind of take you from zero to hero. And...
Amr Awadallah: Exactly. If you knew what that is, then this would be easy to create startups. The hardest part of creating startups is exactly that what you just mentioned right now. That is the hardest. Yeah.
Jake Aaron Villarreal: You talked about ease of use. Talk to us a little bit about the product and what makes it so easy to use.
Amr Awadallah: Yeah. So, uh to to I I alluded to this briefly earlier to build a uh system that delivers on trusted generative AI, you have to build that something called the rag platform, which stands for retrieval augmented generation. And in a nutshell, uh what this does is uh what's called open book machine learning versus closed book machine learning. So what I mean by that h go back to high school and remember if you're taking a physics exam let's say physics was one of your favorite topics and you were taking a physics exam you studied the physics you studied the books you took the courses and everything and you're now doing a closed book exam right so you're trying to solve the exam and it's closed book you cannot look at the book and you see a question that you completely understand 100% understand the question but as you are answering it you discover that oh I forgot the formula for force equals mass multiply by acceleration uh but I needed to solve this question.
So what you what you do or at least I did that myself because I was an A+ student. I always want to get a good grade. I would make up the formula. I would literally just make up a formula and say let's assume this is the formula and this is how I would solve this question if this was the formula but I just can't remember it right now. Uh that is hallucination. That's exactly what hallucination is. It's it's you trying to fill in the blanks because you couldn't remember everything about what you have studied. And that's the exact same thing that happens with large language models. So with large language models, there's a certain amount of memory that they have. They can't remember everything that because we train them on so much data not for the purpose of memorization. We train them for the purpose of understanding. Right? The same way that we go to school, we go to school, the primary reason why we go to school is to evolve our thinking abilities and our problem solving abilities. The memorization is a secondary effect that is also useful. And the large language models have the same problem. So they they're very they're trained on a lot of data. They're very good at understanding, very good at reasoning, but their uh memorization is off by about 2% to 10%. Right? Where they have to make up stuff and that's make up stuff where is where the problems come from.
So retrieval augmented generation is instead about open book large language models. Meaning the large language model will answer the the the question just like you would answer the physics question. But whenever you have any problems memorizing, we're going to give you the book and you can look up the book and say, "Okay, force equals mass multiply by acceleration." and then you'll give me the perfect response without having to make up something. So that's what retrieval augment generation is about. Implementing that as a platform beginning to end is extremely hard. It's very easy to prototype, very easy to make a sample, but very hard to do it as a production ready system that is accurate, that is secure, that is explainable. And uh that's exactly what we have built for our customers is instead of you going and struggling with open source and grabbing pieces from here and pieces from there and trying to build all of this platform beginning to end using the Lego pieces which is the hard part. We give you a pre-built solution. It's a box that does the job very very well. You have an API on one end where you feed in your data. You have an API on the other end where you issue your questions or prompts and everything else just works for you automatically. the security works, the accuracy works, the explanity works, etc., etc.
Jake Aaron Villarreal: That's perfectly well explained. God, that's incredible. Wow. Hopefully, you'll have a lot more customers after this. I know I'll be I'll be sharing this with our port our portfolio. Um, what's the biggest challenge you're facing today as a company?
Amr Awadallah: Too much noise in the market. So, I'll tell you this is this market, this GenAI market is moving at a speed I've never seen in my life. Like uh I thought the big data wars were were were insane wars like when we were starting Cloudera there was like 20 other companies trying to do the same thing of which only a few survived at the end and we merged with one of them called Horton works but in the case of Vectara oh my god there's like many many other companies claiming they can do things when they know nothing about the problem they just claim oh we can do this we can do that when they don't know how to solve the accuracy they don't know how to solve the security they don't but they claim it anyway so that creates noise in the market which requires us to spend extra effort educating the customers on what matters in Geni and what does not matter. I think that's a problem in this year uh because of the fluff. There's a lot of fluff and like like I'm seeing a lot of sim I'm seeing a lot of similarity between this year and 1999. 1999 was the bubble years for the internet where there was massive investments in anything that was anything around it. Like my first company, you you were mentioning my first company was acquired by Yahoo. We were actually acquired in 2000. We were one year old, five people, but at the time that we were getting acquired, there was 20 other companies doing the exact same thing. Almost none of them exist today. All of them failed. They existed because there was a froth in terms of investment. So there is a froth in terms of investment happening in GenAI right now which is creating some of that bubble like there is a bubble effect. Like I have to be honest, I'm seeing a bubble form and and and only the strong players like Vectara and others will be able to survive that bubble and in fact they will be much stronger coming out of the bubble. Exactly like what happened with the internet bubble. Facebook, Google, uh Snapchat, all of these were born out of that amazing uh internet bubble. The iPhone itself.
Jake Aaron Villarreal: Yeah. Yeah. No, you're right. I mean, we've waited 24 years for the next big transformation. We know it's here. We know it. We know it's AI and it is transforming industries which is incredible. But you're right. I mean there's 1,000 to 2,000 AI startups launching a week right now. And we know that because you know all the companies we support currently are are AI companies and...
Amr Awadallah: Wow.
Jake Aaron Villarreal: 100%. Wow. 100%. We've been in business for 13 years. Yeah. So we it just it's it's exactly like the .com boom and I went through that too. I was at Oracle and we saw we didn't I didn't know we were in a bubble at the moment that happened but this um I don't know if you'd say it's a bubble. It's certainly like the beginning of the growth spurt and there's going to be a lot of winners, a lot of losers and a lot of in between acquisition.
Amr Awadallah: There will be some correction. I I I believe so myself. Uh but at the same time it's necessary. This is a necessary step in in the evolution of technology. like we have seen it there's many bubbles that happened in the past by the way I mean of course the internet one was really really big one the one we had around 2008 was another big one uh and they you see the companies that survived them and come out out of them they become stronger like out of 2008 we got Cloudera out of 2008 we got Databricks we got uh Snowflake there's MongoDB there's some really really strong companies that came out of that crash of 2008 so um yes we should be afraid of crashes crashes do create some financial losses. I I don't I hate that. I don't like that. But I also understand from experience that it's a healthy stepping stone in the in the bigger arc of where we're evolving this industry.
Jake Aaron Villarreal: Yeah. What's on the road map as you look through to 2025? We're halfway through the year now. What are you excited about?
Amr Awadallah: Yeah. So, this year this year is the is the year of AI assistants and next year is going to be the year of AI agents or agentic AI as you might hear it. So AI assistants they help us but they don't take action on our behalf right so they are there to give us an answer hey dear radiologist here is the answer for how you can configure your ultrasound machine or here lawyer here is a suggested contract that you can write or here Jake here is a suggested candidate that you might want to talk to and send an email to source next year uh when we get the accuracy to be almost close to 100% which is required for we will now start having an a a massive onslaught of AI agents. AI agents are things that are able to actually take actions on our behalf. So not just give us an answer and we have to uh act on that answer. No, they will act on the answer directly. They will reach out to the candidate directly on your behalf, Jake, and try to source them, right? Once they find that candidate is a good hit, they will directly submit the the lawsuits on behalf of a lawyer. they will directly send a customer support message or read it back to a customer instead of a customer support rep. They will uh directly apply the configuration for an ultrasound scan for an average human not just for a for a for a radiologist.
So that's the that will happen that that is really and that's the that's the really the the the most exciting thing about GenAI actually the reason why I think this is going to be a massive massive market is uh every single application around us every single device around us whether consumer or enterprise it doesn't matter or business will be powered by GenAI they will have capabilities that will allow us to use them verbally so let me give you a very layman example imagine you're using photo editing software to fix a great picture that you took you were out on the beach. There was an amazing sunset behind you, blue water, blue skies, green Hawaiian trees, and you take a picture of your family and it is the best picture ever. And but you go back home and you know, oh, one of your kids actually your son had his finger up his nose in the picture. My son does that to me all the time. So, so, so, but the picture is amazing, right? So, you want to fix it. So, today to do it with photo editing software, you have to go and look at a tutorial in YouTube and maybe read the manual. within a year once this agentic stuff is figured out the UI the user interface we have with applications will become uh verbal but you will just tell the application uh hey dear photo editing software please uh remove my uh son's finger from his nose and the the software will just do that for you it will read its own manual it will figure out how to do it and we'll do it for you right so that that would be true around us everywhere our car our cars sometimes they show us a red icon on the tableau that something is broken in the car right and today when you see that red icon. They're always very crypt cryptic. I don't know who made these icons. You have to open the manual of the car, which is like 500 pages. You have to find that icon to find what's wrong. Again, within a year from now, two years from now, you'll be able to just ask your car, "What's wrong with you?" And the car will read its own manual. We'll look at its own diagnostics and will tell you, "Hey, dear human, or actually more likely, hey, lazy human, please go change my oil. I've been screaming about it for the last two months." So, that this is really the future that we're seeing. This is why I'm so excited. Like, I think this is going to be amazing.
Jake Aaron Villarreal: Yeah. Well, we're seeing the transformation happen already. Uh God, next year if that's the case, we're we're also we're going to be really seeing this agents do the work for us is, you know, it's going to be really exciting.
Amr Awadallah: And I would say with us instead of for us like again, I don't want to scare people that we're going to lose our jobs. They the agents are going to help us become 10 to 100 times more productive. And the advice I give to people, there's people who are scared. The advice I give to people is learn how to use the technology. You need to learn how can you use these AI assistants and AI agents regardless of your discipline. You need to learn how to use it to make yourself more productive and more efficient because the people who are going to lose their jobs are the people that can't use these tools. Like it's very similar to when the IT revolution came with laptops and desktops. The people that knew how and learned how to leverage laptops and desktops, they kept their jobs and the people that did not lost their jobs. So the same thing will happen again with this wave and that's the number one advice I can give to you. Don't shy away from AI tooling. Embrace AI tooling. That's your savior in the future.
Jake Aaron Villarreal: I love that. Well, on that note, I'd like to thank you so much for coming on the show and sharing your story and for all the listeners for listening. Uh it means a lot to me that you spent your time with us today.
Amr Awadallah: My pleasure.
Jake Aaron Villarreal: Awesome. Really, really fun to have you on here. Uh, I'm your host, Jake Aaron Villarreal, signing off for now, but can't wait to catch up with you all in the next episode. Until then, take care. If you like what we're doing, don't forget to subscribe, leave a review on Apple Podcast or wherever you listen, and follow us on YouTube, where we go behind the scenes to learn what it takes to be a startup founder.
.jpg)
.webp)
.webp)
.webp)
.webp)
