328 Googlers Facebook Should Poach

There has been a persistent rumor that Facebook is building a search engine. Facebook is already famous for targeting Google talent, so given that Google is the search engine, it wouldn’t be surprising if Facebook looked to Mountain View to build its search team. Google might have to upgrade those gold handcuffs to platinum if it wants to keep its best and brightest. Now let’s suppose the rumors are true and Facebook wants to target Google’s search talent. How would they go about it?

The Challenge - Finding a Search Team

It won’t be easy, even for Facebook who has the natural advantage of their enormous social graph to slice and dice. Finding brilliant engineers and scientists with a specialized skillset is a little trickier than finding a friend to recommend. This isn’t an algorithm issue, it’s a data - or lack of it - issue. Given their limited numbers, lack of self-promotion and employers’ incentive to hide them, it isn’t always easy to find computer science PhDs. You might identify a Java programmer easily enough, but finding someone who specializes in massively parallel machine learning is non-trivial.

Facebook recruiters would likely take a traditional approach: set a bounty and wade through the thousands of Googler-turned-Facebookies referrals. A company without Facebook’s home-court advantage might turn to headhunters, whose traditional approach would be to trawl LinkedIn and Facebook accounts of Googlers, question Stanford professors and, of course, do lots and lots of Google searches.

Not only are these strategies tedious to implement, but there are limits to these approaches when looking for someone as specific as a search quality engineer because:

- Google has close to 10,000 current and former engineers and creating a dossier for each requires thousands of hours and is not scalable.
- Most Google engineers list ambiguous titles such as “Senior Staff Engineer” and do not
disclose the projects they work on until after they leave.
- Public profiles usually present the most optimistic picture of the author, which means hyperbole is the norm.

With those limitations in mind, we took on the challenge of seeing if we could use our international database of patents to find the right engineers faster and with more accuracy.

Using Patent Data to Build Facebook’s Search Team

We set ourselves a target of identifying 100+ Google engineers who 1) are top search engineers, 2) have a financial interest in joining a pre-IPO company, and 3) are young and hungry. We limited ourselves to using only our patent and company data, Hive queries, a few custom MapReduces and a healthy dose of trial-and-error.

Step 1. Finding the Obvious

First we identified patents assigned to Google. This is trickier than we first thought. Finding the patents assigned to “Google Inc”, its official company name, was easy. However, Google Inc is not the only entity that holds Google’s patents. To find Google’s full patent portfolio we collected a list of subsidiaries and acquisitions, 98 to be exact. Before we could find those companies’ patents, we first needed to back out their official company name as it would appear on a patent record. The old school way of doing that would be to manually search Wikipedia, Google and TechCrunch to weed through all similarly named companies.

We didn’t like the sound of that, so we opted for running the list against our database of 25 million companies. We used a variety of keyword roots to do the name matching. For names like “Aardvark”, we got a couple hundred matches, which we narrowed down using more specific keyword filters. We ended up with a list of 199 official entity names of potential Google controlled companies.

Next we queried our patent database to find all the patents assigned to Google and its 199 potential offspring. We got a result of 25,315 patents.

Then we tallied all the inventors associated with each of these patents, which, again, was not quite as simple as we’d hoped. The initial output gave us 9,147 inventors. However, reversed names, middle initials, misspellings and other name variations made it hard to pin down individuals. Using de-duping methods we reduced the original 9,147 inventors to 7,177 inventors.

Using this method we were able to create a simple ranking of Google’s most influential inventors. We used each inventor’s patent count and how many times their patents were cited by other patents as proxies for influence. The results are what most people would expect, with distinguished engineers like Jeff Dean, Shumeet Baluja and Sumit Agarwal near the top of the list. That’s a good start - we’ve stated the obvious - but that list is not yet really all that useful for Facebook. The list includes top engineers, not top search engineers.

Step 2. Searching for Search

To deal with the search vs. generalist problem we filtered out non-search patents via keyword matching. That narrowed the list of relevant inventions to 1,254. There are 485 people associated with these patents. To further narrow the list, we searched for inventors who appear on 10 or more patents, which revealed 134 people. The Top 10 by number of patents are:

Patents - Name
84 - Simon Tong
76 - Lawrence Stephen
66 - Benedict Gomes
48 - Guha Ramanathan
46 - Amit Singhal
44 - Glen Jeh
44 - Haveliwala Taher
42 - Adam Smith
42 - Sergey Brin
42 - Kamvar Sepandar

At this point we have a high-quality and manageable list of people working on search at Google. The top names are indeed world-class search experts. Unfortunately, they are a little too world-class. The top of this list is a Who’s Who of early Google employees, who are not likely to entertain Facebook’s overtures. Their minds, hearts and wallets are wedded to GOOG, especially Sergey Brin.

Thus, that list is still not all that useful for Facebook. We needed to find out who on the list might actually be willing to join Facebook.

Step 3. Getting Ageist

To identify engineers who might be less financially and emotionally committed to Google we looked for those who are relatively new to Google. The best target is someone at or just past their four year anniversary - the typical vesting period for Googlers. Incentive re-ups aside, choosing this stage of Googlers ensures their best equity days are behind them and that they joined Google after the IPO.

With this in mind we removed anyone who holds a patent - search or otherwise - earlier than 2008.

The Results

Based purely on our patent and company data mixed with some algorithms, the Top 10 Googlers for Facebook’s search team of the future are:

Patents - Name: Title, Expertise
12 - Maureen Heymans: Technical Lead, Social Search
12 - Hui Tan: Engineer, Social Search
12 - Alexandre Kojoukhov: Engineer, Social Search
10 - Othar Hansson: Technical Lead, Google Instant
06 - Alexander John Komoroske: Product Manager, Google Squared, Search
06 - Kacholia Varun: Technical Lead, YouTube / Video Search
06 - Tomislav Nad: Search Quality Engineer, Discussion Search
06 - Vida Ha: Search Quality Engineer, Mobile Web Search; Technical Lead, Google speech recognition
06 - Roberto Bayardo: Google Research, Data Mining and Privacy(!) Expert
06 - Hu Ning: Head of Technology, Mobile Search Quality (has left Google)

Facebook might not have heard of any of these people - we certainly hadn’t - but after a few quick searches, it appears they are exactly the kind of Googlers who Facebook should be targeting. It’s even a healthy mix of engineers, researchers and product people.

The full list has 328 potential candidates and, despite the post title, we’re not going to list them here. Facebook doesn’t need our help.

By expanding this list to include people from Bing, Yahoo, Baidu, etc. we could create an even richer recruiting lead list. By applying our rankings for patent and inventor importance we can apply even more accuracy to the rankings.

Google search engineers was a fairly easy group to find. The same methods can be used to help:

- Intel to discover the top Chinese scientists working on rare earth mineral technologies.
- Twitter to open a London office and find local talent.
- A new hedge fund to identify the top Goldman Sachs quants working on credit default index swaps.

We currently perform custom investigations and make the results from our analyses available as dynamic web-based reports. To learn more contact us at [email protected]


We investigate and acquire datasets from around the world using automated data retrieval and "deep web" crawling methods.


Our analysts use our existing library of proprietary ETL tools to return clean and structured data in any format your business requires.


Raw data is rarely good enough by itself. We use advanced natural language and machine learning methods to extract entities, from people to addresses, so you can find information, not just data.


At over 200 million records, 1 billion entities and relationships and 88 countries, our existing dataset of companies, people, intellectual property, legal and financial filings make for powerful supplements to your existing data.


Sales | Support | Press | Employment | Partners

US: +1 302-566-5993
Hong Kong: +852 3693-1524
Fax: +1 866-594-4383