Ad

Thursday, March 7, 2019

Lessons from 200+ Data Science Interviews

Follow us for beginner friendly tutorials. We will also publish a better quality version of this article on medium.

Notes from Metis webinar of the same title with my own commentary and opinions. Metis is a data science training bootcamp with technical portfolio projects. Metis also offer bootcamp prep courses: python prep for example.

At his previous work Capital One, he started the machine learning powerhouse team that grew from 2 to 80, lots of experience hiring machine learnist and data scientists.

Understand the hiring process and hiring funnel
Employer: Define roles, build a funnel, attract applicants, interview & hire, pro tips

Define Roles
Roles: Data Scientist, Data Engineer, Machine Learnist,
Data Engineer may be a software engineer who codes, and is ready to tackle a large complex dataset, doing ETLs and transformations. Massage data ready for model building.
Data Scientists are ready to handle cleaner data, ready for doing analysis.

Machine learning engineer, a job title that is more senior, a sweet marriage of data engineer and data scientist. JAVA C++, can explain algorithms, KNN, k means.

8:35 In finance, the job descriptions, roles, and skill definitions are well defined, even mandated. Requirements of experiences are set and even strict. Startup role definition tend to be loose, even chaotic.

Set internal expectations. Set expectations for people who conduct interviews. For hiring, it's better to set expectations right to attract the right candidates. Result in smoother experience for candidates. Candidate also feels that the experience is tailored. Google onsite panel is assembled based on candidate strengths and interests.

12:00 data analyst may know SQL but may not know a programming language.
Data analyst, data engineer ETL infrastructure may also need to know data visualization.

Building a funnel and attract talents
13:40
Generally, with funnel, one starts with a large number of people and quickly decreases to a small number of people - monotonically decreasing.

The goal is to build a funnel, attract people to the funnel, and optimizing it. Comment: It is an important startup growth, product management technique.

Application Funnel
Funnel does not really differ by company size.
Entry point: sourcer, referral, cold applications. Sourcer will actively reach out to candidates, but it is a quick process. They won't spend more than 1 minute at your profile.

If you have a person to follow up with, you are already further down the funnel.

Cold application can result in a pool of thousands, or tens of thousands applications.

Referrals are much deeper in the funnel. Even "half way there already". At a meetup. Someone they know may be hiring even if they are not directly hiring.

Non technical Phone screen : pulse check, culture check, is this person generally agreeable, broad skill check with a recruiter, how is their communication skills. Check if this person knows the company language.

17:00
Technical Phone Screen: perhaps everyone's least favorite part, culture check, little CS problem, data science problem algorithms, "coarse filter" for does candidate have enough skills to justify on-site interviews. 4-6 hours of engineering time is valuable so it's best not to waste on candidates that are not ready. Cracking the coding interview book.


On-site : 4-8 hours, a proper day, including technical and non technical interviews. Even a post-on-site sometimes. Discuss candidates feedback, make offer, expectations. Comment: I heard that Microsoft sometimes do this, the day of the interview.

Offer includes salary, compensation, starting date.

First day:  be nice if there's a small celebration.

Sometimes there's a take home. For the presenter : "If someone asks me to do 8 hours of work for free. I will just say no. "

"Just make sure you have a good pipeline. Constantly get people through. With well defined steps." - Presenter on building a great application funnel.

Presenter's preferred method of finding best candidate: organic means of finding candidates. Conferences of relevant topics (shows that they are committed and passionate if they are spending time on a Thursday night for hours learning about a topic), meetups, speaker reception, exchange business cards, rolodex? Existing contacts, networks.

3 comments:

  1. I really enjoy simply reading all of your weblogs. Simply wanted to inform you that you have people like me who appreciate your work. Definitely a great post. Hats off to you! The information that you have provided is very helpful. https://vnmachine.com

    ReplyDelete

Machine Learning Workflow

Data cleaning Missing data Outlier Others: duplicates, typos, special characters Strategy for missing data: imputation, mean, median...