Over the years, my work contributed to many parts of LinkedIn's data & and personalization stack. Through AI & personalization, my work added $xxx million in incremental revenue while improving advertiser ROI.
A few highlights were:
* DuaLip, an extreme-scale linear programming solver that was used to improve matching in internet marketplaces. This was used across the company in matching, (a) job seekers and recruiters (hiring), (b) content creators and engaged audiences (advertising, news feed, courses), (c) senders and receivers of professional connection requests (people you may know).
* Developed an incremental learning framework to allow personalization models to be retrained in near realtime. The system was architected to train ~x million models while keeping the model freshness in the order of minutes.
* Graph & representation learning to embed parts of LinkedIn’s economic graph. Learned representations improved candidate search quality for the largest line of business.
* Made algorithmic & systems contributions that enabled the org to scale experimentation, improved data quality, and reduce cost to serve.
No comments:
Post a Comment