You're Doing the "Two Pizza-team" Wrong, ML Best Practices, Leadership in Hypergrowth, and More
Featured posts from Google, Grab, Facebook, and More
As usual, I've compiled a list of my favorite discoveries from the previous month. I hope this edition inspires you to think creatively. If you missed the last one, you could catch up by clicking here.
If you like this newsletter, please forward it to others who might be interested.
To ensure faster support time for chat users, Grab needed to help their agents become more effective. This post describes how Grab's engineering team created an ML model that provides contextual suggestions by leveraging several internal data sources to improve chat agent efficiency. I especially enjoy their methodical approach to finding a solution to the problem.
I came across this article that explains the difference between a director of engineering and a VP of Engineering as one advances in her career. Although each company's criteria can vary slightly, this post explains the differences and how to prepare for it.
“The ‘two-pizza team’ paradigm has become really popular in the context of organizing software teams. But have we lost the original intent of the ‘two-pizza teams’ (aka autonomous teams)? A team is autonomous when it delivers value to the customer independently”. This post shares details on why the so-called autonomous team in some agile environments is not actually autonomous and may have missed the point.
“Do machine learning like the great engineer you are, not like the great machine learning expert you aren’t.” This guide from Google’s blog contains 43 best practices for ML engineering and is intended to assist those with a limited understanding of machine learning in gaining access to Google's machine learning best practices.
Software engineers often wrongly assume that becoming a successful engineer entails becoming an excellent coder. Knowing how to code will indeed allow you to get stuff done and move quickly. However, once you reach a position of leadership, the amount of responsibility you bear skyrockets. And performing well entails more than just coding; as your responsibilities grow, it will no longer be about what you can do but what you will allow others to do. This is one of the reasons I enjoyed reading this article.
“In other large-scale machine learning domains, such as natural language processing and computer vision, a number of strategies have been applied to amortize the effort of learning over multiple skills. However, because robots collect their own data, robotic skill learning presents a unique set of opportunities and challenges.” This post on the GoogleAI blog introduces two new advances for robotic RL at scale, MT-Opt. MT-Opt introduces a scalable data-collection mechanism that is used to collect over 800,000 episodes of various tasks on real robots and demonstrates a successful application of multitask RL that yields ~3x average improvement over baseline.
“Hypergrowth is the dream of all startups. But once you hit hypergrowth, it can quickly turn into a nightmare. Everything is in constant flux, and no one has an instruction manual”. Here is a good read from Ákos Kapui, VP of Engineering at Shapr3D, about managing a company in hypergrowth, doubling employee count every year.
“The future of AI is in creating systems that can learn directly from whatever information they’re given — whether it’s text, images, or another type of data — without relying on carefully curated and labeled data sets to teach them how to recognize objects in a photo, interpret a block of text, or perform any of the countless other tasks that we ask it to.” Recently, Facebook introduced SEER, a new high-performance computer vision system developed by leveraging self-supervised learning and can learn from any collection of digital images without requiring researchers to curate the collection and label each object.
You can also support me and my work by buying me a coffee. Thank you ❤️