Why you want to work in machine learning at Facebook, by Facebook

Facing reports that a growing number of young engineers are no longer aspiring to work at Facebook, the social networking giant went on the offensive last week. In a livestream on its career page, Facebook hosted a conversation with four of its machine learning engineers to answer questions and talk up why prospective developers should want to join the company.

Perhaps unsurprisingly, considering all the recent public backlash, two of the four engineers were managers working on projects aimed at ridding Facebook of fake users, trolls and harmful content – some of the main areas of concern for prospective engineers considering joining the company. Below are the main arguments for working at Facebook, and particularly as a machine learning engineer.

Seemingly unlimited freedom

The four engineers often came back to the same benefit of working as a machine learning engineer at Facebook. “There is no real code of ownership,” said Darya, a software engineer who works in London. “You can modify whatever you want; you can work on something that you are passionate about and is a good idea and no one would stop you to try and experiment with things.”

Engineers have the freedom to move around the company, change teams and work on different projects as they see fit. “It’s a great way to develop your career without leaving the company,” said Daniel, a manager in charge of using machine learning to remove fake accounts. During the hiring process, new engineers aren’t assigned to a specific team or project. They can try out different teams and choose the one they prefer. “If you don’t like [a project], you can just pass.”

Facebook’s architecture

The engineers also talked up one of the benefits of working for a large tech company: the depth of Facebook’s architectural backbone. The productivity tools available to machine learning engineers at Facebook “allow you to do things that would normally take weeks in a matter of days,” said Fabrizio, a research scientist who works in applied machine learning. Experimenting with machine learning models “is literally pressing a button,” he said.

Self-inspired innovation

Touching on the first point on autonomy, the group said that machine learning is all about experimentation and coming up with your own novel ideas. Employees decide on their own how much time they want to spend doing research compared to engineering. “People propose crazy ideas all the time,” Daniel said. While most teams have a project manager, “people are pretty equal – engineers get a lot of say.”

Your work touches millions

Like many other machine learning experts, at least one engineer on the panel started his career in academia, where the work may affect only a few hundred people. “In academia, you focus on looking for problems. At Facebook, you struggle to pick the most interesting problems to work on – there are so many to choose from,” Fabrizio said. “You’re like a kid in a candy shop.”

Much of the conversation around machine learning focused on this idea in particular – fixing problems. Part of the reason Facebook has seen an apparently dip in interest from prospective engineers is that, unlike startups, much of the core innovation has already been completed. Some students who still have Facebook at the top of their list told Vault in a recent survey that helping to “fix” the company was part of their motivation.

You don’t have to have a background in machine learning

While it’s assumedly more difficult, the group said you don’t necessarily need to have specific machine learning experience to be hired as part of the team. You can work with experienced engineers, start learning features and new languages, and eventually become an independent machine learning engineer. However, “if you want to contribute in meaningful way,” developing a deep knowledge of math basics like algebra and calculus is necessary, said Fabrizio, who has a PhD. “Otherwise you are doing something that you may not fully understand.”

Also, you probably need to know Python and C++, the two programming languages most often used in machine learning, at least at Facebook. Experience working with PHP and Haskell – which runs Facebook’s entire spam-fighting system – also seems helpful.

Projects are never “done”

After a viewer asked how machine learning engineers find a new project once one is completed, the four panelists all started laughing simultaneously. “Completed?” one panelist responded. Machine learning engineers tend to work on “ambitious” projects that can span multiple years, progressing toward a goal “that may not even be visible” yet, said the host.

Main takeaway

While the webinar certainly focused on machine learning in particular, one general point was hammered home consistently: Facebook believes it offers engineers unprecedented freedom and autonomy to work on what they want to build their own career. That’s their sales pitch. Plus, you get to fix stuff.

Have a confidential story, tip, or comment you’d like to share? Contact: btuttle@efinancialcareers.com

Bear with us if you leave a comment at the bottom of this article: all our comments are moderated by actual human beings. Sometimes these humans might be asleep, or away from their desks, so it may take a while for your comment to appear. Eventually it will – unless it’s offensive or libelous (in which case it won’t).