Ville Miettinen – Microtask 1 of 2

"The cool thing about being an entrepreneur is that you get to do everything." Microtask enables efficient global trading of task-based digital labor for enterprises.

Matthew: Hi, this is Matthew Wise with FounderLY. We empower entrepreneurs to have a voice and share their story with the world, enabling others to learn about building products and starting companies. It’s with great pleasure today that I’m here with Ville Miettinen, who is the founder and CEO of Microtask. Microtask is a real-time platform for distributed work. With that said, Ville, we’d love for you to give our audience a brief bio.

Ville: I’ve been an entrepreneur my whole life, so when I got out of high school in ’94, I went to the University of Helsinki to study – I lived in Finland back then – to study computer science. That same fall, we founded our first graphics technology company, which was called Hybrid Graphics. At Hybrid what we did was build some of the world’s very first 3D engines, so that’s the software that’s needed for rendering the 3D graphics for a game. We licensed that to a number of games companies here in the U.S. as well as in the UK.

And then, in the late ’90s, we started specializing in the space of massively multi-player games, and we built a technology called Umbra, which is still being used in a lot of MMO games. We spun that company off a few years later, and that tech is used for rendering and managing extremely complex, huge 3D environments, with hundreds of thousands or millions of simultaneous players. That’s the technology that was used back then in games like Star Wars Galaxies or Everquest and so forth.

Then in 2001, we were contacted by Nokia, this big Finnish, or at least they used to be this big Finnish cell phone manufacturer, and they asked us to do a research project, a pure R & D project to find out whether it’s possible to bring 3D graphics into mobile phones. You’ve got to remember at that time, the specs of the mobile phones were very different than what the current generation iPhone 4′s are, so the machine we had was a 60 MHz RN7 processor, with a 96 x 65 pixel display with black and white, not even gray scale, but black and white.

What we did was we took the open GL rendering standard and we cut it down a bit and then we crammed that down into 50 kilobytes of RN code so that it was placed on the actual ROM of their phone. And we were able to do some of the very first 3D demos running on open GL on a phone and at that point we realized that if we are able to pull this off and something cool like this results, maybe the other mobile phone manufacturers would be interested in something like this.

And obviously, this goes where? We’re all thinking about how to bring 3D graphics into cell phones. And the way that happens in an industry like the cell phone industry is that big companies, the handset manufacturers, the big operators and so forth, they get together, and they organize themselves as standardization committees. So, there was this standardization group called Khronos. Being foreign, we joined in. Khronos developed a standard called open GL ES. That’s the standard that’s still being used, for example, in the iPhone or in Android. Most of the 3D graphics is probably being done through that API.

So, we participated in that process, and we built the very first software implementation of open GL ES, and then subsequently first implementations of a number of different Java rendering standards like JSR 184, JSR 239, and some others. And then, we licensed this to the handset vendors. Nokia used the technology. Samsung used it. Ericsson used it.

And then, it was picked up by various operating systems on Java VM vendors so it drifted into various consumer electronics and eventually also even into the automotive industry. So there were car manufacturers who were replacing their physical dashboards with digital, or the physical instruments in their dashboards with digital dashboards where they needed everything rendered and they used the tech.

I was one of the founders. I was the CTO, originally. I’m a programmer, assembler, [inaudible 04:50] hacker. I’ve been doing that since I was a toddler, but roughly ten years ago I started drifting over to the dark side, doing technical marketing and technical sales. A lot of evangelism, standardization work and things like that. When the mobile graphics market really started growing, like 2005 or 2006, what happened then was what usually happens when a tech sector starts growing, there is massive consolidation.

So, there was this two or three month acquisition spree during which the big players in the industry just acquired all the smaller companies. We were the first one to be acquired, by NVIDIA and we became NVIDIA Finland. Immediately afterwards, the other small players were acquired by ATI and ARM [SP] and the other big boys who tried to break into the mobile graphics industry. I stayed on board at NVIDIA for about a year and a half, learned everything I know to this day about large corporations and left at the end of ’07 to start looking for what’s the next new, new thing I want to work on?

Matthew: What makes Microtask unique? Who’s it for, and why are you so passionate about it?

Ville: The concept of Microtask really originated back in 2007, after I left NVIDIA and I was traveling around the world. I do a lot of photography. I’m a semi-professional photographer as well. I was on a trip to South America, and when I got back from there I had about 10,000 new photos on my memory cards. On my hard drive, I have about 150,000 photos, which haven’t been classified properly and they haven’t been tagged. I started thinking about this tagging problem. How do I get tags applied to the photos so I can do searches, so that a particular photo would have the information like, two girls, mountain, forest, blue sky, daytime, and so forth.

Obviously, once you start thinking about something like this you quickly realize that it cannot be done with current generation computer vision and artificial intelligence. So, you need human intelligence brought in. You somehow need to outsource this to people. My initial idea was that my next company is going to be building a product that would be doing photo tagging integrated into Aperture, and Lightroom and Photoshop, so you could just outsource the photo tagging for people somewhere out there.

The more I started looking at the problem, I realized, first of all, that this alone wouldn’t be a really big business and the real interesting stuff, the thing that I became passionate about, was the underlying architecture that’s needed in order to build a service like this. So, how to route tasks to people all over the world through the Internet? How to reach the people, how to compensate for their work whether in money or other terms? How to ensure the “quality” of the results? How to manage the whole process?

So, I got really interested in this, and I started reading more and more and doing research about crowd sourcing as in realizing what the big implications of this kind of micro work-based crowd sourcing would be for the world. If we would be able to convince big companies, big enterprises to slowly start switching over from traditional outsourcing into this kind of micro work, it would allow large enterprises who have big, very labor intensive projects that are traditionally outsourced to business process outsourcing companies, we allow them to outsource that in a distributed fashion so that it’s done by untrained, unsupervised people all over the Internet.

Primarily right now, we’re focusing really on one kind of work and that is text recognition. Text recognition is used, for example, when a company has a very large number of documents, where they need the data entry to be done for the documents, convert it from paper form into digital form. And that requires a combination of machine intelligence, things like optical character recognition algorithms, but it also requires human intelligence, especially when dealing with things like handwriting or difficult to analyze cases.

Our current customers are companies like insurance companies, who have a lot of paper forms, banks, survey companies who are doing traditional surveys. We also work with quite a few national libraries and national archives, who have extremely large volumes of printed material which they have microfilmed and scanned subsequently, but they haven’t done the conversion of the text so that it could be properly searched. One thing that would make an impact is to be able to bring actual real, properly paying work to people, let’s say, in rural India and so forth.

Internet connectivity allows us to move work from one place to another. Now, we have very inexpensive terminal devices. So, not only things like the OL PCs and super cheap PCs, but also cell phones. So, people can be reached through cell phones. Now, the only question is really figuring out how to bring in a lot of work through this kind of a system, and I think that’s the one thing I was super passionate about. If we’re able to pull this thing off, either Microtask as a company or the wider cloud labor industry, we’re going to make a pretty big impact on a lot of people’s lives.

Matthew: Who are your co-founders? How did you meet and what talents and skills were you looking for?

Ville: The founder team is actually four people the founder team. Harri, who’s essentially our COO. I’ve worked with him for the last 18 years, so Hybrid, the startup I was telling you about earlier and some others startups we set together. So, we know each other very well. Then, we have Panu Wilska, who’s our VP of sales. We’ve worked together now, I think, for nine years. We head hunted him back in ’03 to Hybrid, where he ran our sales operation there, and he’s got a history in building sales operations here in the U.S.

Then, the fourth co-founder is Otto Chrons. He’s our CTO. He’s got a very similar background as myself. He’s been an entrepreneur for the last 15 or 16 years, set up his own engineering outsourcing company in Tampere in Finland. He had about 150 engineers in his team, and they were mainly doing work for Nokia. Then, back in ’06, he sold the company to Sasken, which is a very big Indian outsourcing company. He was the Director of Technology over there. Otto, after he left Sasken, he started thinking about really this micro work related issues, and we just semi-randomly met one day for lunch. We were both discussing about what we were thinking about doing next and it kind of clicked that we were thinking about the same thing.

Matthew: From idea to product launch, how long did it take and when did you actually launch Microtask?

Ville: The way it happened was that back in 2008, I was still traveling and doing various angel investments in Finland. After the NVIDIA exit, we made some money out of it, so we were investing that money onwards. So, I started thinking about the whole Microtask thing and I contacted my co-founders-to-be. They were still working at NVIDIA. I told them about the ideas we were having and told them, guys, you’ve got to leave NVIDIA. You’ve got to stop working for the man and you’ve got to go for another start up. And they left at the end of ’08 and we founded the company in, I think, January of ’09.

We spent the first year, and in retrospect I’ve got to admit that that’s way too long in the kind of environment, but we spent the whole first year in just building a first version of the platform, and then we went and talked with, I think, 150 companies out there. Because we had this real-time task broker in technology but we didn’t know what to do with it. So, we talked with 150 companies, and asked these guys, hey guys, what do you do? What does your company do? What are the very labor intensive processes you have? Could they be formulated as these small interchangeable microtasks? And then we built, I think, close to ten different prototypes.

We tried our hands on a number of different things, but could this be used for speech transcription? Could this be used for distributed translation? Could this be used for face recognition? Could this be used for product categorization for e-commerce companies and so forth. We built these little prototypes and tried out different things and eventually we settled on text recognition for a number of different reasons. But that process took about a year.


FounderLY Photos

			FounderLY posted a photo:	Future of Education - FounderLY			FounderLY posted a photo:	Future of Education - FounderLY			FounderLY posted a photo:	Future of Education - FounderLY			FounderLY posted a photo:	Future of Education - FounderLY			FounderLY posted a photo:	Future of Education - FounderLY			FounderLY posted a photo:	Future of Education - FounderLY

Press Mentions #1

Press Mentions #2