How blockchain is helping sell unused cpu cycles and creating opportunities
Interviewed by Christophe Limpalair on 04/22/2018
Blockchain is probably the single hottest technology right now, and our guest today is using it for something you might not expect (hint: it’s not another cryptocurrency). A repeat guest, Jack Levin shares a new product that his company, Nventify, has been working on to help companies with terabytes of data process their documents and images more efficiently. Instead of spinning up thousands of servers to run these tasks and paying the major cloud providers a lot of money to do it, these tasks are run in the background by people like you and me, who are willing to sell unused CPU cycles on our own computers.
How can you apply blockchain outside of Cryptocurrency? Find out today with our repeat guest Jack Levin, who is the founder and CEO of Nventify.
Jack, give us an overview of your background. What are some of the roles that you've filled in the past, and what are you working on today at Nventify?
[00:00:54] Jack: Thanks for having me again, always a pleasure. Where I come from is basically an IT background. Officially I'm a computer science major, but I haven't really been a software developer in any way. I have spent a lot of time developing for computer systems and working architectures.
I came into The Valley in ‘97, but my first job in—I think it’s summer of ‘97—I spent about one year working there. The company closed very quickly. It was a startup—it didn't work out.
The next job, which I guess I would call my official job was Google, and that's with Sergey at Stanford in ‘99, and they only had seven people in the company. I spent a few minutes talking to him and he said, "Yes, we're building a search engine, it's going to be awesome. Do you need the job?" And I said, "Yes. Sure, I can stop by and check you guys out." The rest is history.
I spent about six years building Google’s architecture from the ground up. It was a lot of fun, I really did not know what it would become. I definitely learned a lot, because when I came in I figured I have some basics and I definitely spent all those six years building on those basics.
Eventually, after Google in 2005, I decided to go ahead and start my own thing so I started this startup called ImageShack , which was essentially an image hosting company and not that dissimilar from Flickr or PhotoBucket. Raised money, hired a bunch of people, essentially had a lot of fun developing that company, and ultimately, once we figured that sort of business that we're in is extremely competitive, the advertising supporting revenue was being flat, we essentially decided to close it down and start a new company called Nventify.
Now Nventify is where I am today, and Nventify is the company behind this product that we call Imagizer. Imagizer is a cloud image transformation solution that allows companies to solve their image manipulation problems. Specifically, Nordstrom.com, if you go to Nordstrom.com and browse their image catalog you're going to see that all of their pictures are nicely sized in mobile and all of the resizing and manipulation and transformation happens by using Imagizer.
We're focusing on large e-commerce companies today, like Macey's, and Home Depot. Also trying to hit all the other big names like Pinterest. Essentially if a company has a lot of images in their image library we would like to offer our services as far as image manipulation, so that's where we are today.
[00:04:04] Christophe: Imagizer, which we cover in the first episode in quite a bit of depth, can process images on the fly and resize them in 25 milliseconds or less. It's a very impressive technology. A fascinating interview, please do check that out.
Jack threw me a major curve ball right before starting the interview. I came in here thinking that we were going to have a different conversation and then all of a sudden, he drops a bomb saying that a lot of the clients they've been working with have lots and lots of images that they're trying to figure out what to do with. Now, they're doing something with blockchain to help solve some of those problems. Tell us more about what you've been doing for the last couple of months.
[00:04:49] Jack: This is an entirely new project, it's very exciting. I am personally very excited about it. It really started out from the factors where Imagizer is a real-time image optimization service, and a system and a layer. Some cases exist where, for example, a client of ours may have up to a billion images and they need to process those images, not necessarily on the fly but they actually need to derive or maybe extract some metadata or tag those images. We're talking about terabytes and terabytes of data. Essentially, it's big data meeting computer vision, and one way to solve problems like this is to build a very, very expensive and large computing cluster.
Like something that you would be able to do on AWS or Google Compute Cloud, it would cost you a lot of money. You would have to hire a lot of engineering resources and essentially build custom software and brand your image processing within that cloud or within the cluster.
The approach that we're trying to develop is essentially we're planning to use blockchain as a distribution and fulfillment layer for the clients that have let's say a ton of images, and let's say that that client actually wants to find locations or people’s faces on all of those pictures. Currently, if they were to run a cluster with let's say 500 servers or even 1000 servers, it might take up to a month to run the servers’ image processing.
Now with blockchain, we can have what generally people would call miners—rather than mining for hashes in cryptocurrency, those miners or workers can receive a piece of code that will have a set of instructions as well as receive this subset of images and run image processing on their spare CPU cycles on their laptops or at home or anywhere else. It's similar to SETI, for example, or folding at home where both of those systems run distributive computing but here we're using blockchain as the backbone that receives the tasks and passes those tasks and subtasks to the willing participants which we call workers or miners.
Again, miners—rather than mining for cryptocurrency—ask to do something useful, like request fulfillment by a common client, essentially. Once they finish their task they get paid by the cryptocurrency or maybe PayPal or everything can be done through our portal.
Two things: one, walk me through how Imagizer fits into that, what part it plays, but before we do that tell me if my analogy is correct, if I'm thinking in the right direction here. You know how with Captcha where you have to type in the letters that you see in the image to confirm that you're not a robot. When you do that it's my understanding that, in the past at least, you were actually helping decipher texts that had been scanned to be able to be added into a digital library, so oftentimes the machines weren't able to decipher if this is an A, B, C or D or et cetera.
By having the people type in the confirmation it was helping those engines figure out, yes okay, this is an A, this is a B, this is a C, et cetera. What you're talking about here if I'm right, is you're essentially removing the need for that human being to do that but at the same time you're paying that person for their computational power, right? Is that kind of the same similar thing?
[00:08:51] Jack: It's even simpler than that. When it comes to computer vision and face detection, pretty much anyone today can take an existing neural network that's been trained to recognize faces. That neural network is not necessarily large but it's been computed, it's smart enough. It has very high accuracy of say, 95%. There's a lot of open source tools that are available. It's actually part of Imagizer today.
With Imagizer you can do that through face detection. What we would do is we would take our Imagizer code with our neural nets, push it around to our workers on their blockchain and have them do the work of Imagizer but do it, not in time of the work like Imagizer works today, but rather receive a very massive task, let's say do a face detection on a hundred thousand images. It may take a couple of hours to do but if you parallelize this task and have 10,000 people working, that's 10,000 computing hours crammed in one hour, right? With which you could get that sort of system launched or get your workers organized within two minutes rather than building out the cluster that runs 1,000 servers.
When it comes to software, or specifically any sort of detection, all of that already been invented. What hasn't been invented is the ability to organize spare CPU cycles from the people who are currently mining for cryptocurrency.
Again, I'm not saying that you can only be a worker if you're mining, you can be a worker if you essentially decide to visit imagizer.com and sign up to be a worker. We'll give you a piece of software that you will run in the background on your laptop that will essentially wait for tasks and you can actually manage how many CPU cores or memory or time you want to rent out to a potential client.
So when you sign up, you can receive those requests and as you receive those requests, it processes it and then sends it back through your engine, to the requester (the client) that's asking for that work to be done. Is that right?
[00:11:37] Jack: Yes, that's exactly right. That's how it's going to work when we launch. The blockchain layer facilitates the transactions between the requester and the worker. There's also a verification layer, for example, you can have workers that may potentially have bad hardware, and so they don't return the proper results. In this case, we would be using a consensus swarm algorithm where some workers will be acting as QA verifiers and some workers will simply be there and do work.
Once you reach positive consensus that this work is actually accurate, you can send it back to the client, and the client will essentially flag it as the task completes, and all of the workers get paid.
So it's not dissimilar to AirBnB. With this, you have some space that you can rent out for people to live for certain period of time. Here, if you have spare CPU capacity, you can totally rent it out by connecting to our blockchain network, and becoming a worker that can receive and do tasks.
Now, when I said face detection, this is only one example, you can do other things, like, for example, you can say, "Here is 100 million PDF documents, we really want to transcribe them in text." You would run something like a neural network that will read and place words in the text files, and those text files will be returned back to the requester.
The other potential tasks would be that if you have very high resolution images and you want to convert them. For example, if all of those images are, let's say, they're TIFF before or PNG or JPEG, but you really want them to be WebP, the latest Google format for web images, you could technically run a task to convert between formats for your complete image library, so you would potentially pay only $100 and get your task done in one hour. No engineering resources required.
Walk me through how that works on that client’s perspective. How do I go, if I'm XYZ customer, and I have those billion images that I want to have some work done against it, how does that work? How does Imagizer fit into that?
[00:14:20] Jack: Okay, so what you would do is, you would go to our portal imagizer.com. You would publish your task for a consumption or work available. There, you can basically say, "I have that many units. This is what I wanted to do with them." We can literally publish what's possible. Like, for example, WebP conversion could be one task, face detection could be another task, image tagging can be another task. So you publish your, essentially source of where your data is which could be Amazon S3 for example.
All of your images are sitting there, they're waiting to be processed, and what our system will do is it will take your tasks and wrap it into a JSON container or kind of work description, then broadcast across our blockchain to all of the available workers. If workers agree that they want to do this particular task, where the reward is high enough. Let's say that I want to be paid $10 an hour, or $5 an hour, or $1 an hour. You, as a worker, can configure your client side to only be connected to jobs that pay that much. If somebody pays 50 cents an hour, you can say I will reject this job.
Once this negotiation completes, you become, as a worker, you could become part of this, part of the peer-to-peer network, part of the swarm, receive your task, get your task done, get paid for it.
Essentially, it's up to a client order request to put up the final dollar amount in totality. They wouldn't necessarily even know how their funds are going to be distributed because you can have a hundred workers working on your task or you may have 10,000 workers working on your task. So our system will handle the funds disbursement and payments in a safe manner through the blockchain.
One of the things that really impressed me the most with Imagizer in the first episode was the fact that you didn't have to manage any infrastructure at all, all you had to do is focus on the software component of it because then a customer of yours who lives on AWS would just spin up an image with your software into it and just pay for the fee of the instance and your software, and then you would take a cut out of that so you never had to deal with any of the infrastructure. Is that still the case? Is that still how it works? Or do you now have to add some infrastructure in order to manage that transaction?
[00:17:17] Jack: Are you talking about specifically blockchain related stuff or Imagizer as it is now?
[00:17:27] Christophe: Both.
[00:17:31] Jack: Okay, so our philosophy always was is that we really wanted to reduce any friction between the clients and the service. Imagizer.com is extremely simple in terms of how you can configure it and how you can pay for it. In other words, we don't lock you in into a contract of any kind, there's no really sales process besides a POC, and the POC is just being a proof of concept where we give you a bunch of credits that you can run on imigizer.com. It's not dissimilar to AWS billing model, where you pay as you go, and that reduces the fact that there's really no any sort of commitment or locking on the client side when using Imagizer.
Going back to the infrastructure and the simplicity, that was a very intentional decision on your part to keep it that simple, whereas some of the other image processing providers have infrastructure to manage and it's a different kind of offering and model. One of the major challenges that product and engineering teams and CEOs struggle with sometimes is keeping that simplicity when it comes to deciding what kinds of features to implement, the messaging, the market that you go after, etc, and so my question is: when you made that conscious decision, correct me if I'm wrong, but that probably even limited the size of the market that you can go after because some of the customers might require for you to have that infrastructure so they might go with another provider that does it instead.
What was the thought process behind limiting that market and instead being hyper-focused on a certain group or a certain market or vertical, what was that decision like, and why did you make that decision?
[00:20:54] Jack: Well, I think the decision came from the fact that we really wanted to study the markets and affect the market, and provide services for the markets that we could have the most impact at. In other words, we decided to go after a commerce market because those companies and those businesses are hyper-competitive between themselves.
They are always looking for performance improvement, where if you just basically look at all the companies that provide images online or access to the images online, they do have the need for image manipulation but they're not as sensitive to the performance as e-commerce companies, for example, would be because it is a well-known fact that if you wait one second or if your webpage is delayed by a second, you actually lose about 7% of your sales in the year.
For large companies like Macy's, Nordstrom, Home Depot, this is a huge hit on their sales. Let's say that we talk about social media apps or social media websites, it is not the same thing for strategy, their users are mostly engaged with the sites in a different way. When it comes to e-commerce, it's all about showing the products in the right kind of format, getting people to buy these products and execute those transactions.
We focused on the largest market that has the largest impact on our bottom line. For example, when we tried to sign up a client, we really wanted to make between half a million to a million dollars a year with that client when it comes to the sales. We're not necessarily interested in 10,000 clients that are going to be paying us $100 a month. That is not what we're focusing on.
By restricting our scope, we are able to tune the product, which is Imagizer, so that it can be really good for that specific market, rather than being very, very scattered across all the features that are possible and try to satisfy everyone but really struggle, pretty much everybody wanting you to do a different thing and pulling you in different directions.
Being a small team of seven people, we made a conscious decision to really focus on e-commerce, specifically on large e-commerce companies so that we can dedicate our resources to working with them specifically, such as like for example, if Nordstrom calls or Macy's calls, we shouldn't really be talking to a thousand other companies that are small, we really should be focusing on those large companies. By being a small team, we're able to really give them full attention.
Did you always plan on going that direction and focusing on larger but fewer customers, or was that a conscious shift of you didn't start in that direction and then you realize through whatever method that that was the direction for you to go after?
[00:24:35] Jack: It wasn't originally the plan. The funny thing about running your company or developing a software service business is that everyone has an ego, and I certainly have an ego, right. I certainly want to think that our service, that our software is needed by absolutely everyone. I wouldn't say it's a very difficult process but it's definitely a process to convince oneself to leave some money or some effort on the table so to speak and focus on this one vertical and be really good at it, but I think it's the right thing to do, egos aside, and because ultimately it changes how your bottom line is affected.
For the first year or so, we dabbled in different markets. We looked at social media sites, we looked at mobile apps, we looked at mobile site development companies. Ultimately, we realized that it is e-commerce that where we can make the most money or the most revenue in the shortest amount of time. It simplifies our operation because once you talk to three e-commerce clients, it is very likely that if you talk to three more, you will speak the same language. In other words, you don't have to explain your product to them in terms that you do not know or you need to develop a different angle for them. It's all pretty much the same thing because they all have very similar problems.
That shift in mindset also required a change in messaging, so whereas before, you might have talked about how Imagizer can help load images a lot faster through the web for your customers, that might speak more to the engineering level, but for the types of customers that you're going after, the engineers may not actually be the decision-makers so instead you have to talk about the reduction in costs, the increases in revenue by loading images faster, and every second of delay causes X amount of revenue loss, and so on and so forth. How did you get to the point when you realized that messaging needed to shift? And how do you recommend that other companies figure that out on their own?
[00:27:07] Jack: I think this can be done preemptively through research. In other words, if you go to Google, obviously. And you type "website performance top 10 things you must do" it is very likely that you're going to see that, "If you want to make more sales, your pages should load quicker, they should organize properly, they should have proper formats, you should be using a CDN, so on and so on and so on." It's not a very difficult process to get there. If you spend time doing research, and just pretend that you're running a company like an e-commerce company.
I have recently met a person who started a shoe company in Utah, and it was very interesting to go through a process of how they got the whole thing up and running. What their competition is and how they think. Definitely, I would say, the first thing one should do is just look at the abundance of information that's available online, go to conferences, talk to people, interact, network, see what kind of problems that the chief revenue officers are waking up thinking at night or if you're a CEO of a smaller e-commerce company -- just ask them a few questions like, "What are you worried about? Is performance important. Are you using a CDN? Why aren't you using a CDN? How many CDNs are you using?" For example, right. "Are you using Amazon Cloud Services or not?"
Just come up with a list of five questions that you would ask a person at the conference and just go through it and record it in your mind. Build that sort of profile, and then you actually know what you probably should look like because you will find patterns and people end up creating this template for you to be able to build a product that is compatible with.
The other part of that is sales. One of the tendencies, especially with technical founders, is to focus a lot more on the product than the sales component of it because they'll think if the product is amazing, it'll speak for itself, it will grow organically, people will refer to it, and what not. Maybe they don't focus as much on sales or they delay hiring a sales team for a while. In fact, I think that's one of the things that you did at ImageShack as well since I quote you told me, "We always thought that this would be a technology company so we didn't invest in sales."
So the first question is, when do you think that is the case? When do you think that it's okay to really pour all of your efforts into the products, delay the sales component, or even don't have any sales component versus focus on both the products, then finds somebody that's more sales oriented.
[00:30:05] Jack: Well, I would say that's an easy question with an easy answer. If your product is bad, it sucks and you know it, no point selling it. You're not going to be successful at selling something that doesn't work very well. You absolutely have to either raise money, hire people, or work on it yourself. Take it to a level where you're satisfied with it. It doesn't need to be a complete package, but whatever it does, it has to do well. If it does image conversion between different formats, that has to do it well. It cannot crap out on certain image formats.
You have to bring a product to the market that you can actually sell. Not rely on essentially thinking that you're working so hard, then you put so much time in it and you have so much passion. None of those things will actually make sales. Wait until your product is ready, hire a sales team, preferably hire a VP of sales that can organize the sales organization under and push to the markets.
How did you find your VP of Sales and how do you measure their success when maybe you don't have that sales background?
We found our VP of Sales at the conference called SaaStr. The way we found him is that we were looking to go to learn more about the software service market, and we saw his picture as a testimonial for SaaStr. I took down his name, and I took down everybody's name, then I checked to see if he or all those people are coming to the conference. We went to the conference, we found him, we pitched Imagizer, and his first reaction was, "Hey, this is just like by piper, my favorite show in image compression and real Silicon Valley thing." He really got interested, and we had a lot of good rapport pretty much right away. Within a few weeks, he was hired at Nventify.
I was going to go back and say, I've definitely seen some crappy products with great sales team do fairly well. Usually, that doesn't last very long though because then you have a competitor that does actually build a better product and they start taking over. Anyway, I digress there. To wrap up more of the wild card question, what is your opinion of Bitcoin right now?
[00:32:54] Jack: It's a loaded question.
[00:32:57] Christophe: [laughs] That's the point.
[00:33:01] Jack: I think about Bitcoin in two different ways. One, Bitcoin is a storage of value that you can trade. I would say 90% of people in the world are excited specifically about that. It's putting money into Bitcoin and having that money grow, multiply, or whatever. That is interesting, it's really speculative and exciting for some people.
The other aspect of Bitcoin is the underlying technology like blockchain. I am very interested in blockchain because it's mature enough, and it's been around for long enough to prove to the world that it's very difficult to circumvent as far as the platform is concerned.
Ethereum is another blockchain platform that makes blockchain more production ready that you can do great things on, like smart contracts. I'm very excited about that, and I'm very excited about building on that platform. I think combining the economic value of Cryptocurrency with the things that you can do in peer-to-peer fashion on the technology end, makes it almost as important as essentially start of internet in 1992.
We are at the level here where blockchain can, not necessary solve problems, but it can open doors into so many interesting venues for technology and economics. I'm very excited in all kind of different businesses just coming up and using this technology. Right now, if you go to IBM, you can actually launch a private enterprise blockchain for yourself using IBM Cloud. You can do the same at Oracle. I see Google and Amazon very quickly developing similar products, where you can just look into your clouds, rent a piece of blockchain, and do a lot of interesting things with it.
I know that you're an adviser, also. Have you invested in any blockchain companies working around blockchain that you can share publicly or are you looking at any companies right now that you think are doing really cool things with blockchain?
[00:35:26] Jack: Not specifically, I haven't really invested in many startups, but I'm definitely looking out for startups that are working with blockchain technology. If there is a startup out there, I would definitely be seriously considering investing. One company that's public called overstock.com, I simply bought their public shares. I think they're one of the companies that are using blockchain or their new token called tZero—I think that's interesting. I wanted to buy their shares just this year to see what would happen. Definitely, I would say if you want to get exposure to blockchain technology as an investor, just buy all the Cloud companies, buy Google, buy Amazon, buy IBM, buy Oracle. Likely, all these companies would benefit from blockchain use.
Jack, thanks so much for your time again for the second episode. It's been a pleasure having you on the show both times. If anybody wants to follow up with you if they have any questions, how can they reach out to you?
Reach out to me via my LinkedIn, or my email address - jack at nventify.
(Question after we stopped recording that was interesting)
How did you learn what to do with all these stuff, are there any blogs you follow, any books, any specific people you follow?
[00:39:38] Jack: There's like few portals out there, one of them is called ICO drafts. Or if type in Google, ICO, or listings. You can get a list of old companies and projects that have had a successful ICO, an ICO meaning Initial Coin Offering, you probably know it by now. You can read their white papers to see what their proposed project was and what they're doing or what they're trying to do with their token. Reading those white papers can actually tell you exactly what they're going after. They spent a lot of time polishing it, it's not anybody's opinion. It's the real project that got funded. If you read 10 of them, you can understand exactly what they're going after and you can learn a lot from their perspective on the market and the blockchain specifically.
Blogs or medium, those could be fun but it would be people's opinions. If you literally get a list of 10 or 15 successful ICO's, download their white papers, go through them, your blockchain IQ will multiply rapidly. That's one way to do this.
Thanks everyone for tuning in. Please, leave a comment below, thank Jack for his time and see you next time. Thank you all for tuning in.
How did this interview help you?
If you learned anything from this interview, please thank our guest for their time.
Oh, and help your followers learn by clicking the Tweet button below :)