Nice interview, but I wish she shared more technical details on what they actually did, what worked and what did not.<p>Especially on bootstrapping a recommendation engine from sparse data set, which is a very common challenge -<p><pre><code> It was definitely hard in the beginning because we had lower density of data.
As more people joined, however, we got increasing insight into how the graph
connects. We were then able to look at several things people have in common.
More and more signals began to appear.
</code></pre>
Would be great to see what data they had, how they prioritized which signals to use and how they evolved their algorithms as more data was generated.