Amazon Interview Question - writing sample
I recently successfully interviewed with Amazon.com and was offered a senior program management position within their San Francisco engineer lab. As part of the interview process they asked me to submit a writing sample based on their question “What was the most inventive moment in your career?”
Below is my answer: You might find it helpful and interesting.
TagCow.com - Image tagging problem solved
One of the most inventive moments in my career came just as I was about to throw in the towel on a start-up I had founded in 2006. We had been trying to bootstrap a subscription based scanning service for people’s non-digitalife ( mail, old photo albums, children’s art, etc.,) but it quickly became clear that our approach wasn't scalable so my co-founders and I were looking for a pivot. One of the byproducts of the prototyping and business process modeling we had done for the scanning service was a collection of 100,000+ digital images from my childhood. One day I was trying to find some specific photos that I had seen during the scanning of my images and realized that at scale you can’t “browse” your images you have to somehow enable search. This idea of enabling one to search their photo collections kept coming up over and over again during the following weeks. I would want to show a friend or family member a photo but they would grow bored watching me “browse” through my giant collection of photos. I was certain that there were services available to annotate one’s digital images with meaningful metadata and thus enabling one to search against their image collection so I immediately set out to identify this service and solve the problem for myself. What I found would lead me to my Eureka moment.
I did some research trying to find some image tagging solutions and was surprised that such a service just didn't exit. What I found was that while image recognition had come a long way it was still a long way off from being good enough for consumers. The problem was that only computer scientists were working on the problem and they were looking for a completely automated solution but I knew instantly that I could solve the image tagging problem for everyone by combining algorithms and humans! I was already an Amazon Web Service customer and was familiar with the Amazon Mechanical Turk service. I hatched a plan to license a set of image recognition algorithms called Alipir from the University of Penn State and use Amazon’s Mechanical Turk service to train the algorithms and augment the automated tags with tags generated by the human workers. I called a meeting with my co-founders and they instantly loved the idea - we had our pivot and on that night in January of 2008, TagCow.com was born.
Three months later we completed development of a beta candidate. I was tasked with marketing the service which I had thought a lot about in the weeks leading up to our initial launch. We had no funding so I knew a massive and extensive consumer marketing plan was not an option. Luckily, I had recently read Malcolm Gladwell’s legendary book, Tipping Point. So the same night we completed our beta I scoured the internet for our mavens. Immediately I stumbled onto a gentleman named, Thomas Hawk, who at the time owned a blog focused on the popular photo sharing site, Flickr. In his bio he indicated that he currently had over a quarter-million digital photos and was aiming to shoot more than a million in his lifetime - I knew I had the perfect beta tester. I sent him a personal invite to join our little beta program. That same night he attached his massive Flickr archive to our tagging service. Within a few hours we had tagged his entire collection of images. He was so impressed that he phoned his friend Michael Arrington, editor of Techcrunch.com, and raved to him about our service. By midnight of that night I received a text from Michael Arrington asking if I was available for an interview. I knew our service wasn't prepared for the Techcrunch effect so I called my co-founders to discuss a strategy. Before we could complete our conversation I received a text from a friend of mine congratulating me for being on the front page of Techcrunch. The headline read “Image tagging problem solved”, he would go on to describe our service as “magic”. In the coming hours we would sign up tens of thousands of users and tag millions of images. Here’s a link to the article (which Michael Arrington would later edit after he realized we were using Mechanical Turk): http://techcrunch.com/2008/03/29/image-recognition-problem-finally-solved-lets-pay-people-to-tag-photos/
We would be featured in Techcrunch two more times in the coming months:
No comments:
Post a Comment