What is it like to be a machine learning software engineer at Facebook?

Answer by Tao Xu:

I joined Facebook in 2009 with an official title: Machine Learning Engineer. I liked that title a lot, mainly because:

  • I love machine learning
  • Its acronym (MLE) also stands for Maximum Likelihood Estimation.

I seriously hope that someday Machine Learning Engineer will be as popular/trendy as Data Scientist.

The job of Machine Learning Engineer is to apply machine learning to practical problems. Most of people with that background were concentrated within a few core ranking teams (for example, Ads, Feed and Search). In my case, I was in Ads Ranking for the first three years and Feed Ranking for my last year at Facebook.

Most of my daily works are around:

  • Developing new ML algorithms/models scalable to FB data, typically we changed the underlying ML algorithm in a major fashion every 1.5 years.
  • Writing increasingly faster runtime for the existing models, so it could score more candidates with lower latency. Those core ML runtime routine is running a few trillion times a day.
  • Developing new features
  • Developing new system infrastructure to store ever more data to extract features at runtime
  • Developing ML infrastructure to make model training easier, faster and more realtime.
  • Training model and started A/B test
  • Staring at A/B test results
  • Writing/running tons of hive and map reduce jobs
  • Slicing and dicing data.
  • On-call, debug data correctness, metrics drop etc.

View Answer on Quora


Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s