Shariq Rizvi, the current Director of Ads at Twitter, was also one of the co-founders of Twitter’s Performance Ads, helping it grow from $0 to billions in revenue.

1. Why did you originally invest in Rover and what excited you about that technology?
2. Do you think that type of technology can be disruptive in the content recommendation space?
3. Do you see any similarities between the innovations you and your team pioneered on the Twitter Ads platform and the potential of Content Recommendations in general?

Barney Pell, former R&D at NASA and early manager at Microsoft, created Powerset, a natural language search engine that Microsoft acquired. Pell was an early architect at Bing.
1. Why did you originally invest in Rover and what excited you about that technology?
I invested in Rover originally because I knew the founders before they started the company and felt they were super smart and hard working. I also liked their push into scalable personalized content recommendation technology using machine learning and viewed this as a challenging and important problem.
2. Do you think that type of technology can be disruptive in the content recommendation space?
I definitely think this type of technology can be disruptive in content recommendation. The entire business is based on the quality of the recommendations made for each user in each specific context. And you don’t always have the full information about the user, context, or content alternatives, so you have to make the most out of whatever data you do have. This is a big data problem where accurate scalable solutions can make all the difference.
3. Do you see any similarities between the A.I innovations you and your team pioneered at Microsoft and the potential of Content Recommendations in general?
The thrust of our innovations with Powerset were about deeply understanding the meaning of the content, down to each individual sentence, and letting the user express their interests in natural language, and making the best match. The Rover technology is not trying to match to individual facts on the one and, and takes advantage of more contextual and user historical data on the other hand.