ISSUE 320   ·   January 26, 2021        

Insight

"Very few techies are trying to do the big hard things."

Many of the large tech companies have started "AI for Good" programs and claim they want to help solve big problems like climate change, sustainability, and social issues. Are they for real? And if so, what challenges do they face besides who's actually going to do the work?
Marc Winkelmann interviewing Clive Thompson

Seven Legal Questions for Data Scientists

From privacy and algorithmic bias to security and third-party dependencies, there are a lot of ways that data can get you in trouble. Thinking through the questions discussed in this article will help ensure that you and your organization stay aligned with the law.
O'Reilly Media | Patrick Hall and Ayoub Ouederni

Sponsored Link

What is a Feature Store?

Feature Stores manage the lifecycle of features that power ML applications. They enable data scientists to build features quickly and reliably, and provision them to production instantly. This post explains what a feature store is and describes its main components.
Tecton

Tutorials, Projects & Opinions

Unsupervised Data Monitoring

In this first post in a new series, Jeremy Stanley shows how relying on rules to monitor data quality doesn't scale and why you should use unsupervised learning instead. The examples here are great and, if you're interested in data quality, this will definitely be a series to watch.
Anomalo | Jeremy Stanley

Building a team of internal R packages

If your organization has started building its own packages, this post by Emily Riederer offers a great approach for designing and tying together an ecosystem of internal tools. Using the jobs-to-be-done framework as a guide, she explores strategies for API design, docs, testing, and more.
Emily Riederer

ML Theory with bad drawings

Nice introduction to machine learning concepts with lots of helpful diagrams along the way.
Windows on Theory | Boaz Barak

Recommending Articles With Help From Our Friends

It seems like a simple problem. Let readers select interests from a list of categories and then send them personalized content. But at scale, that quickly gets complicated. Here's how and what the New York Times learned from a different approach using machine learning.
NYT Open | Joyce Xu

AMAX: Experience the Fastest Solutions for AI

Take a test drive on AMAX’s GPU-accelerated server solutions with the latest NVIDIA Ampere GPU architecture to kick start your most demanding analytics, inference, or training workloads with access to AI Expertise for a seamless testing integration. Experience the most powerful end-to-end AI & HPC data center platform now!
// sponsored

Career

Building a data science startup

Interested in turning a data science side project into a small startup? Listen to this discussion with four current founders to learn about some ripe opportunities and key skills you need.
TalkPython['Podcast']

Wanted: Data Scientists with Technical Brilliance AND Business Sense

Technical skills are important but when it comes to making big impacts, business intuition is key. In this post, Alok Gupta describes how entrepreneurial data scientists at DoorDash approach problems differently and get more done, faster.
DoorDash Engineering

Data Viz

Density Plots

Scatterplots and line charts often encounter performance and overplotting issues as the amount of data increases. Density plots are an alternative that can work better for larger volumes of data.  Here's how they work, including a library and interactive examples to play with.
Oberservable | Twitter Open Source

Data Elixir is curated and maintained by Lon Riesberg. For full-text search of prior issues, visit Data Elixir's Search Page. If you have suggestions or questions for the newsletter, just reply back to this email.

Sign up to get Data Elixir's  data science newsletter in your Inbox >>