ISSUE 403 · September 6, 2022
Join the Data Elixir Talent Collective
The Data Elixir Talent Collective is a reverse job board where top companies apply to you. Members control all communication, so you won't get the noise that's typical on other recruiting channels. Choose to be anonymous or public and get matched with opportunities that fit your specific interests; be it a career move, more pay, remote work, etc.
The Collective is off to a great start with 75 members in its first week.
This is a free resource but membership is limited to professionals with 3+ years experience in data science, analytics, ML, visualization, and related fields. The plan is to create the highest signal-to-noise hiring resource for roles in the data ecosystem. For more info, apply here.
Insights
Down the Semantic Rabbit Hole
The purpose of a semantic layer is to close the gap between the “business language” and the “data language” and offer a unified and consistent view of the business. This deep dive into the topic is one of the more complete explainers I’ve come across.
JP Monteiro
Be Good-argument-Driven, Not Data-driven
When you begin to favor *bad* arguments that involve data over *good* arguments that don’t, then you’ll notice things start to go awry. Richard Marmorstein offers a reminder for us data folks to consider foundational elements of the argument itself before leaning on metrics.
Richard Marmorstein
Tutorials, Projects & Opinions
Communicating A/B Test Results with Ratios and Uncertainty Intervals
If you work with A/B tests frequently then the question of how to communicate uncertainty has likely come to mind. This argument for presenting A/B test tests for conversion rates as ratios, with uncertainty intervals around those ratios makes a lot of sense to me.
Harlan D. Harris
How Instacart Uses Machine Learning-Driven Autocomplete to Help People Fill Their Carts
Search is a big deal for eCommerce platforms like Instacart. This post describes how the team generates and ranks query suggestions in autocomplete and how this shapes a user’s search behavior — translating into larger basket sizes.
Tech at Instacart | Esther Vasiete
Measuring Downstream Impact on Social Networks by Using an Attribution Framework
It’s not the easiest task to quantify the impact that a certain action has in social networks because those actions have secondary impacts across other members. Here, LinkedIn introduces a practical framework for running experiments in these types of environments.
LinkedIn Engineering | Qiannan Yin, Derek Koh, and Jenny Wu
Bayesian Age/Period/Cohort Models in Python with PyMC
PyMC can be a powerful tool in the right hands. Austin Rochford does a good job of showcasing this through Bayesian APC models, including lots of code snippets for getting you ready to tackle the inferential challenges these models pose.
Austin Rochford
Creating a Real-Time Feature Store with Materialize
Why is everyone talking about feature stores, yet so few are (successfully) implementing them? In this post, we walk you through how Materialize can be used to create a real-time feature store for a fraud detection use case
// sponsored
Code & Tools
R Shiny Now Available in Python
For the uninitiated, Shiny is a package that makes it easy to build interactive web applications and dashboards. It was previously limited to the R programming language but recently, the creators of Shiny have announced Shiny for Python as well!
Dario Radečić
DocQuery: Document Query Engine
DocQuery is a Python library that makes it easy to extract information from documents. Works with semi-structured and unstructured documents, such as PDFs, scanned images, etc. Simply point DocQuery at one or more documents and specify a question you want to ask.
GitHub | Impira
Data Visualization
The Magic of Matplotlib Stylesheets
The “out of the box” visualizations in Matplotlib aren’t fantastic. Luckily, stylesheets can level up your plots in a very accessible manner. This overview takes you from fairly uninspiring visualizations to something much more compelling.
datafantic | Robert Ritz
 |