Presented By
O’Reilly + Cloudera
Make Data Work
March 25-28, 2019
San Francisco, CA
Please log in

Toward deep and representation learning for talent search at LinkedIn

Gungor Polatkan (LinkedIn)
4:20pm5:00pm Wednesday, March 27, 2019
Average rating: ****.
(4.33, 3 ratings)

Who is this presentation for?

  • Research scientists, machine learning engineers, applied scientists, and data scientists



Prerequisite knowledge

  • A basic understanding of machine learning and neural networks

What you'll learn

  • Learn how LinkedIn applies and deploys deep ranking models to its large-scale production search systems
  • See how to use these approaches in your own systems


Talent search and recommendation systems at LinkedIn strive to match potential candidates to the hiring needs of a recruiter or a hiring manager expressed in terms of a search query or a job posting. Recent work in this domain has mainly focused on linear models, which do not take complex relationships between features into account, as well as ensemble tree models, which introduce nonlinearity but are still insufficient for exploring all the potential feature interactions and strictly separate feature generation from modeling.

Gungor Polatkan shares the results of the company’s deployment of deep learning models on a real-world production system serving 500M+ users through LinkedIn Recruiter. Key contributions include:

  • Learning semantic representations of sparse entities within the talent search domain, such as recruiter ids, candidate ids, and skill entity ids, for which LinkedIn utilizes neural network models that take advantage of the LinkedIn Economic Graph
  • Deep models for learning recruiter engagement and candidate response in talent search applications

You’ll explore learning to rank approaches applied to deep models and the benefits for the talent search use case. Gungor ends by presenting offline and online evaluation results for LinkedIn’s talent search and recommendation systems and discusses potential challenges along the path to a full deep model architecture.

The challenges and approaches discussed in this session are applicable to any multifaceted search engine.

Photo of Gungor Polatkan

Gungor Polatkan


Gungor Polatkan is a machine learning expert and engineering leader with experience in building massive-scale distributed data pipelines serving personalized content at LinkedIn and Twitter. Most recently, he led the design and implementation of the AI backend for LinkedIn Learning and ramped up the recommendation engine from scratch to hyperpersonalized models learning billions of coefficients for 500M+ users. He deployed some of the first deep ranking models for search verticals at LinkedIn improving talent search. He enjoys leading teams, mentoring engineers, and fostering a culture of technical rigor and craftsmanship while iterating fast. Previously, he worked in several notable applied research groups at Twitter, Princeton, Google, MERL, and UC Berkeley. He has published and refereed papers at top-tier ML and AI venues, such as UAI, ICML, and PAMI.