Swedish Startup Gavagai - exclusive interview

Interview with Swedish Startup Gavagai co-founder Jussi Karlgren.
Scandinavian startups exclusive

"Sudden Success Comes After 15 years of work"
It took less than 6 months for four year old Gavagai to secure their first investor, and they did not even have to pitch. A simple  friendly conversation at a scientific event at SICS, and a demo of their revolutionary Ethersource, got them the funding they needed.
Of course, they also had an amazing product, a hungry market, and all the experience one could ask for.
A good example of swaying luck in your favor? Perhaps.
The formula of success is the current hot topic in startups circles. The factor list is almost arbitrary. Some look for the easy way out, stressing the importance of marketing, selling to investors, and hoping for that get-rich-quick appeal of public stock exchange listings. Those interested in this abstract "make money" approach are quite often cursed with hating their jobs, customers, and supporters. It is sadly the exit that interests them most, and perhaps therefore they rarely make it to that magic door.

It seems that doing what you love is a lot more fun, and as facebook and many other startups before it manifest, also a very effective way of insuring success.

Jussi is a linguist obsessed with finding the interconnection of meaning (here Jussi humbly comments that it is his co-founder Magnus Sahlgren that stands behind Ethersource. "He is the maniac!" Magnus developed his theory under  Jussis scientific tutelage). The  name "Gavagai" is inspired by a thought experiment by the Harvard philosopher Willard Van Orman Quine. Being an expert in a field that is currently very much at the heart of next generation information tech is hardly enough to found a successful startup.  Both founders can also code.

And don't forget the skill needed to listen to advice that comes from selected partners. Of course, you have to be able to find the right advisors as well, and be able to make the right desicions under pressure. In short, all those years working on what they loved created an "impossible to fail" platform.

What is it that you do, exactly?

On the website, it states "Gavagai develops automated and scalable methods for retrieving actionable intelligence from dynamic data." I can hear the groans from my readers now -"Oh please, not another keyword sniffer". Not quite: their tool, Ethersource,finds patterns and hidden context. It does it largely in RAM, in a rather smart  way, making it quicker than traditional analyst tools.

And oh boy, does the market sound sexy. "Ethersource is currently used to track market mood with respect to tradable assets; to assesslevel of security threat to targets at risk; and to monitor consumer sentiment with respect to brands. " Ok, stocks, security, and brand management. This is real money we are talking about. And Gavagai has real customers, paying for that data. Ethersource is a self-learning system. When we were discussing current Elections in Russia, Jussi said that it would take a day for the system to learn Russian (just point it at a public source of cross-referenced data, like Gavagai's favorite Wikipedia) - enough to start tracking mentions of some concept of interest without needing exact search string matching - the system generalises from mentions to the conceptual level!
Throw in awe and some hot trendy words like "neural net", "self-learning", "Bow Down Before Your Killer Robot Overlords".
I would have named the system SkyNeuroNet, but that is just me.

Ethersource is designed to be Scalable, Multilingual, Robust, Complete.  Sounds nice. The live system tests seem to match the claims.

It works amazingly well, amazingly quickly. The scientific background behind is quite sound, and is most certainly not lacking, just look at the amazing publication list. Unlike most of my interviews, I had to ask for less information, as semantic analysis at the level it is being done by Gavagai is extremely complex. The bastard system also managed to get a higher TOEFL and ESL synonym tests for English score than little old me, a well-read gentleman and a native speaker.
If you asked me what Gavagai was about, I would have to say "finding meaning before it is evident". I would certainly not want to write about another "keywords trend mining" application. It is yesterdays news, and the current incarnation google trends feels like the altavista of pattern recognition.
Gavagai looks for concepts, not keywords. But I don't want to get into the theory of it, suffice to say that I like how it applies.
You can find a good practica illustration of Gavagai's approach on their blog, check out how the "abstract" theory translates into better trend monitoring with the illustration of this blog post on Influenza surveillance.

So far, poiticians, governmental development bodies and trend analysts have been quite happy to use the competitive analytical advantage Gavagai provides.

Ether Ether through the wall, who's the smartest of them all?

Gavagai is growing quickly, with a core team of 12, add key advisors. There is a package services solution in the works, geared towards customers in the marketing, stock and trend analysis sectors. Gavagai is about to appoint their first product sales manager, so far, customers have been knocking on their door by themselves.

A very interesting case study: Ten months ago, Taimour Abdulwahab set off a bomb in central Stockholm. The image below illustrates the weak signal Violence Propensity Index (VPI) related to Sweden that Gavagai picked up in the Swedish blogosphere between August 1, and December 21, 2010. Note that the suicide bombing in Stockholm is preceded by approximately 30 days of increased volatility and rising VPI levels and a significant spike on the December 8, that is, three days prior to the event.

Gavagai comments:  "Now, we’re not claiming that the bloggers did it. What we do say, however, is that the attitude towards a given subject, as expressed in on-line social media, may well reflect the attitudes at large in a population, including people who are about to take action and externalize their opinions. The VPI is a means to detect (weak) signals of violent chatter, and as such, may facilitate an early warning pertaining to targets at risk." Yes, big data, the noosphere, global consciousness manifest in weak links. Call it what you will, the link shows up again and again, check their blog. Interesting to see this phenomenon in action, it is being mined by other startups as well, for example Recorded Future

Funding
It is still the first round of funding for Gavagai, and very little of the company stock has been given away for funding. The details of funding recieved so far are not yet public.
"If I had to do it all over again, Ethersource would have been much better placed at IBM or Google from the start. But unfortunately, large companies are not able to scale rapidly, this is perhaps a future holy grail to find", smiles Jussi.
"We are open to a buy-out, but we insist on retaining productive control of our technology".
"My Advice To Startups? Passion. Be honest to yourself. And manage your Risks. As a researcher, I would suggest a founder with a scientific background to use her training as an advantage. Have a theoretical base, have a hypothethis. Focus, do something well, and concentrate on risk management. If we want to take a more philosophical approach, you should have no questions about doing what you love. You have to follow your passion all the way."

Doing just that certainly seems to lead to spectacular results, at least for Gavagai.  It is rare to find a company, let a lone a start-up with such a defined culture, essence. They know who they are, what they do, and why they do it.  I find it deeply meaningful that the wikipedia page about Swedish Institute of Computer Science and it's Notable Spin-Off Companies list Gavagai as one of the projects, yet when you click on Gavagai, you are taken to the description of Willard Van Orman Quine's experiment. That experiment perhaps defines what the company does a lot better than a press-release. They are on a search for meaning. And perhaps that search is the only meaning that truly is.

The Gavagai Team
November 2011
Magnus Sahlgren, born 1973, Chief Scientist and Partner
Dr. Sahlgren, one of the founding partners of Gavagai, holds a PhD in computational linguistics from Stockholm University. Sahlgren’s dissertation was awarded the prize for the most prominent scholarly achievement of 2006 at the Stockholm University Faculty of Humanities. He has worked on computational models of meaning since 2000 at the Swedish Institute of Computer Science (SICS) and Stockholm University.
Jussi Karlgren, born 1965, CEO and Partner
Prof. Karlgren one of the founding partners of Gavagai, holds a PhD in computational linguistics, a PhLic in Computer and Systems Sciences, both from Stockholm University, and is an adjoint professor of language engineering at Helsinki University. He has worked with information access-related language technology since 1991 at the Swedish Institute of Computer Science (SICS), at Xerox PARC, at New York University, and at Yahoo! Research in Barcelona.
Fredrik Espinoza, born 1969, Chief Technical Officer and Partner
Dr. Espinoza was most recently the CTO and chief architect of Reco.se. Previous experience includes founding and managing a wireless applications startup and a position as senior researcher at the Swedish Institute of Computer Science. He holds a PhD in Computer and Systems Science from Stockholm University.
Fredrik Olsson, born 1971, Chief Data Officer and Partner
Dr. Olsson holds a PhD in computational linguistics from the University of Gothenburg, and a PhLic in the same subject from Uppsala University. He has worked with language technology since 1998 at the Swedish Institute of Computer Science, and most recently, at Recorded Future.
Ola Hamfors, born 1975, System Architect and Partner
Mr. Hamfors holds a MSc in Engineering Science from the Royal Institute of Technology in Stockholm. For the last 10 years he has worked as a senior consultant for one of the leading Java consultancy firms in Sweden. Most recently he has been enterprise systems architect at Nasdaq/OMX, Ericsson, and The Swedish Police. He also has research experience from the Swedish Institute of Computer Science (SICS).