#158 Part 1- How do we get data scientists going in 8 hours with Nitish Mathew Global Head of Data Engineering at Afterpay
When Global Head of Engineering, Nitish Mathew looks back at his journey over the last four years with Afterpay, he describes it as a phenomenal ride and a tremendous learning journey. They have had to reinvent themselves three times in a very real way, getting rid of their first and second platform and building with the view of not assuming that this is for the next 10 years, but building with the view that they will need to get rid of this in two to three years.
Nitish says when a data scientist joins a company they should be able to query a data lake or data warehouse like any library. We need to help users find information quickly so they can actually get the information that they want to do their research immediately. It needs to be organised so people know how to find what they need.
We end part one with a discussion on how organisations should use coaching to get the best results. Ultimately, leadership's role is to get the right outcomes and support people with coaching, technology, decision making and resources. He believes the goal is not to build the fastest product. The goal is to make sure your data scientist colleagues are happy. And for them to be happy, you need to make sure that she or he is able to do their job fast, which involves giving proper easy to use performance tooling, giving good data, and then making sure that on a daily basis, it works.
Enjoy the episode!
Quotes
“My role is to actually coach the engineers and the managers across the multiple teams to make those decisions themselves.”
“Science is something and the scientific method has been invoked for millennia. What are the technologies? What are they thinking in science that we can adopt? If you take any scientists, they go through a process of research, experimentation and putting things or making things now we call it production education.”
“Frankly, if we do our job properly with empathy to users, and to enable them to find things fast, we probably can actually do away with a lot of the additional constructs we have had to put on top of that like data resume metadata and all those things.”
“Everybody wants queries to come back in 10 seconds. Now, that is a problem that can be solved only with really good technology, and really good education. “
“People need to talk to each other, we need to get technology, and we need to actually coach our colleagues on how to use these things properly so that they can actually get the best outcomes. “
“Success is the goal that I should be setting for my engineering colleagues. Your goal is not to build the fastest product, your goal is if your data scientist colleagues are happy.”
“15 years ago, there was a famous picture of a world leader saying “mission accomplished” on top of an aircraft carrier. Yes, the mission is still on. Launching a data platform and saying Mission accomplished - no, it's just the start. Your mission on a day to day basis is, is your data scientist colleague happy.”
“Can I actually provide a framework where the data scientists themselves can add a lot of tests that actually are one every time when you want to click the button to push to production so that they feel confident? Right? Nobody wants to do a bad job?”
“Ultimately you can write a lot of tests, but there's really nothing like for certain models, actually putting it out there and getting actual feedback, you have no idea what reality is. And then faithfully being able to put that to a small set of people. And that's a pattern that web application people have been using for decades.”
“I think leadership's role is to get the outcomes right and then support people with coaching, technology, decision making with resources. I think everything else should just work out on its own.”
Thanks to our sponsor:
And as always, we appreciate your Reviews, Follows, Likes, Shares and Ratings. It really helps new data scientists find us.