One of the most interesting companies at DEMO 2006 was certainly www.kosmix.com. Kosmix may be best described as having a 'Horizontal approach' to vertical search. Kosmix has a simple and clean interface that lets you search like in any other search engine. The results however are presented with a pretty helpful categories structure.

So when searching for 'New York' you get results sorted in categories like:

Hotels (9,959)
Museums (1,346)
User Reviews (1,002)
Travel Guides (1,326)
Restaurants (101)
Kid-Friendly (32)
Adventure Travel (182)
Travelogues (41)
Travel Magazines (25)

Founder Anand Rajaraman was so nice to answer some questions for us:

Q: What did you do before Kosmix?

A: My co-founder Venky Harainarayan and I ran Cambrian Ventures, an early-stage venture capital firm. Our goal at Cambrian was to invest in disruptive technologies. Our investments included Neoteris (acquired by Juniper), Kaltix (acquired by Google), Fatlens (product search), Mobissimo (air/hotel/rental car search), and Efficient Frontier (optimizing adwords campaigns).

Venky and I have worked together since we met as grad students at Stanford. We started Junglee, a comparison shopping pioneer, in 1996.
Junglee was bought by Amazon in 1998.

Q: Why do we need categorized search?

A: Existing search engines work well when you look for a single web page e.g., searches such as "united airlines" or "stanford university".
However, we often are looking for the answer to something a lot more open-ended e.g., suppose you are planning a trip to Iceland. A linear list of pages that contain the word "iceland" is not very useful. That's when our categories make it much easier to get at the information you need.

Categorization helps you search less and find more.

Q: Why is Kosmix better then Vivisimo?

A: Vivisimo uses a technology called "clustering". They are a meta-search engine that clusters the results of another search engine. There are a couple of problems with clustering:

(1) The clusters that are created are machine-generated by looking at the frequently occuring words in the top matches. Those words are often non-intuitive to people, or make no sense in the context of the query.
E.g., when you search for "brain cancer", one of the clusters is called "died" and another is called "heart". Kosmix categories are always intuitive to people.

(2) Clusty (Vivisimo's consumer-facing engine) looks at only the top results to create their clusters, so the clusters don't capture all information in the "tail" of the search results. For example, when you search for a disease such as "diabetes", you don't see clusters for Alternative Medicine or Clinical Trials. Kosmix is able to surface this information because we have our own index of the web.

Q: What is the secret sauce in Kosmix's search technology?

A: We call it "algorithmic web categorization", the ability to categorize the web and search results into meaningful buckets.

Q: How many searches are currently performed at Kosmix every day? How many will be performed in future?

A: Not as many as the Google or Yahoo, but a meaningful number given that we just launched a couple of weeks ago and are growing steadily.

Q: How do you actually market the site - through Google?

A: The best marketing is word-of-mouth, through satisfied users who tell their friends. We also work with publishers in the health, travel, and politics categories to distribute our search across their websites to a targeted audience. For example, we are powering the web search for the HealthCentral Network, which has over fifty health-related websites including Dr. Koop and HealthCentral.com.

Q: Where does Kosmix develop software? Why there?

A: We are based in Mountain View, California. There is just a tremendous pool of talent to draw from in Silicon Valley when you're starting a company.

Q: What was the defining moment when pitching VCs?

A: We didn't formally pitch VCs. We showed our prototype and described our plans to a few people we know and trust, and some of them got really excited and wanted to be a part of what we're are doing.

Q: What is your TOP lesson learnt as an entrepreneur?

A: Be careful when building the team and hire only A+ people.