The Data Exchange Podcast: Sean Taylor on how data science and the role of data scientists have changed over the years.
Subscribe: Apple • Android • Spotify • Stitcher • Google • RSS.
This week our managing editor Jenn Webb and I speak with Sean Taylor, Data Science Manager at Lyft. Sean was previously a research scientist and manager at Facebook where he was instrumental in the creation and release of Prophet, a very popular open source library for time-series forecasting.
This conversation focused on the evolution of data science. Some of the topics we discussed include:
- How data science and the role of data scientists have changed over the years
- Recruiting, mentoring, and managing data scientists
- Sean’s recent decision to cut back on management duties so he can carve out time to work on core data science problems at Lyft
- Responsible AI
- Tools for hyperparameter tuning and optimization (Ray Tune; Facebook Ax), time-series forecasting (Prophet; Greykite), and more.
So I remember when I first started, literally running a Hive query successfully felt like a good day at the office. So you know counting things and having good logging and infrastructure was really hard for a while. I think we’re getting to the point where a lot of that stuff is better these days and we get to focus more on applications and problems and fancier models. So that trend is sort of allowed us to get gradually more sophisticated and take on harder problems and more ambitious kinds of projects.The other big theme, I think, that I’ve noticed is this specialization – it used to be that “data scientist” was a very generic job title. You would kind of do a lot of different things: as a data scientist you’d be responsible for all kinds of things like data quality, all the way through to building applications and, you know, even like a web front end for your, for your models and things like that.
Now I think people have a little bit more purview for specific things. Like we have data engineering roles that are becoming more important that kind of help set the stage for the data science work that other other people are going to do. We have these machine learning engineer roles that are more focused on integrating models with a product experience. People are even further specializing: more statistical versus more AI or kind of fancier model based roles.
Download a complete transcript of this episode by filling out the form below:
- A video version of this conversation is available on our YouTube channel
- Reza Hosseini and Albert Chen: “Building a flexible, intuitive, and fast forecasting library”
- “Model Monitoring Enables Robust Machine Learning Applications”
- Paolo Cremonesi and Maurizio Ferrari Dacrema: “Questioning the Efficacy of Neural Recommendation Systems”
- Sercan Arik: “Neural Models for Tabular Data”
- Nicolas Hohn: “Reinforcement Learning For the Win”
- Rumman Chowdury: “Responsible AI meets Reality”
2021 NLP Survey
The 2021 NLP Industry Survey is now open and we need your help. The survey takes less than 5 minutes to fill out and in exchange we’ll send you a copy of the survey results + a FREE pass to the 2021 NLP Summit (a virtual conference slated for October).
[Image: Langkawi Sky Bridge from pxfuel.]