Hiring Relevance Science Engineers at LinkedIn
March 7, 2016
I get asked a lot of questions by my friends about relevance science and engineering at LinkedIn. Why has relevance science become so important in the Internet industry? What does being a relevance engineer mean at LinkedIn? What relevance products are we building? How are we building those? What are the skills required to build those? And most of all, how does one become a relevance engineer at LinkedIn? In this post, I will provide answers to these questions.
LinkedIn operates the world’s largest professional network on the Internet with more than 400 million members in over 200 countries and territories. Our members contribute an enormous amount of information to this network - making it an enormous source of information. The scale of this information is such that a member could find it overwhelming to effectively perform any desired task in a reasonable amount of time. For example, if a member is looking for a job, she could spend weeks browsing through our database of millions of job postings to find an opening of interest. The same task could have been carried out effortlessly if we inferred the skills and interests of the member through explicit or implicit cues and recommended relevant jobs.
This is the role of a relevance engineer. We build products and services that empower professionals to become productive and successful. Building a relevance product or service requires several important skills, all of which are tested in our hiring process.
What we look for in job candidates
We will start with a phone interview to assess your technical foundation. We assess this by evaluating you on your competence in two primary areas. First, an understanding of core theoretical data mining concepts, such as probability distributions, estimation, clustering and classification, to name a few. Second, the ability to write clean, modular code for simple problems such as shuffling an array, or merging two sorted arrays, or the like.
Assuming you receive a positive interview feedback, you will be brought to one of our vibrant corporate campuses where you will get an opportunity to meet some of your prospective teammates, and experience the innovative culture of our fast growing company. Depending on your expertise and experience, you will then go through four to six interview modules. Each module is designed to test a skill required when building a relevance product, from inception to production.
Our product design module will test your ability to design a viable relevance system for a real-world relevance problem, such as designing a deal recommendation system. It is expected that you come up with a high-level layout of the system to facilitate a discussion around various aspects of the design, such as data analysis, feature extraction, model training and evaluation, distributed computation, deployment, and A/B testing. After this, the remaining interview modules will test your ability to implement the various components of a real world relevance system.
At LinkedIn, we have really large volumes of highly structured data. The first step towards unlocking the potential of this data is data analysis and visualization. Our data programming module is intended to assess your fluency in translating data analytics ideas such as sampling, summarization, and visualization into code. You will be asked to implement one such idea, and based on your experience, you may also be asked to extend your implementation to a distributed computing environment.
Once we have analyzed the data, the next step is to pick a suitable machine learning algorithm and train a model. Our data mining module is intended to assess your in-depth understanding of theoretical data mining and machine learning. You might be asked to discuss the objective, parameter inference, optimization, and evaluation of a learning algorithm. Based on your experience, you may also be tested on large scale machine learning.
The final step is deployment of the model into a production service. Our coding and algorithms module is intended to assess your ability to write correct, readable, and fully-tested code. You will be given a problem of moderate algorithmic complexity and asked to come up with an efficient algorithm and implementation for solving the problem at hand.
If you are a specialist within a related area of interest to LinkedIn, such as natural language processing, information retrieval, security, advanced statistics or optimization (to name a few), then we might add an additional interview module or a tech talk for you to demonstrate your skill in that area.
Apart from these technically rigorous interviews, you will also interview with one of the managers in the relevance organization, who will try to find a team for you based on our requirements and your training, experience, and ambitions.
At LinkedIn, you will have the opportunity to develop innovative technologies at massive scale to create economic opportunity for every member of the global workforce. You will contribute to this mission by conducting applied research targeted at solving some of the hardest technical challenges in the industry related to data-driven products such as search, social and economic graph, computational advertising, and the feed. You will be able to truly transform your career in one of the most compelling R&D setups in the Internet industry with our world-class data infrastructure, supportive peer groups, and innovative culture. We look forward to welcoming you to our LinkedIn family.