You are viewing content from a past/completed QCon

Presentation: Probabilistic programming for software engineers

Track: Modern CS in the Real World

Location: Pacific DEKJ

Duration: 5:25pm - 6:15pm

Day of week: Tuesday

Slides: Download Slides

Share this on:

This presentation is now available to view on

Watch video with transcript


Big data — and the neural networks we use to make sense of it — have taken the industry by storm! But what might be falling through the cracks?

In this talk, I’ll introduce you to the world of Probabilistic Programming Languages, and why it’s something that the industry should care about today. Some of the most pressing problems in machine learning concern accuracy, interpretability, and reliability of our models, and PPLs, as they’re called, offer a compelling means to handle all of these at once. At the intersection of programming languages and machine learning, PPLs build upon centuries of Bayesian Statistics knowledge to offer predictions qualified with uncertainty estimates. But they also lean on techniques from the Programming Languages space to make this accessible to ordinary developers. And best of all, PPL models can be trained generically!

In addition to introducing you to PPLs and Bayesian Statistics, this talk will give a sneak preview of how we’re advancing probabilistic programming at Facebook, as well as some of the big problems we’ve used it to solve.

Speaker: Michael Tingley

Engineering Manager @Facebook

I am the engineering manager for Facebook’s Probabilistic Programming Languages team. Our goal is to platformize Bayesian modeling and analysis within Facebook, and invest energy and research into cutting-edge techniques that rely on compiler-driven analyses in order to advance the performance and reliability universal Bayesian inference. We have a diverse team, ranging from Programming Language experts, modeling veterans, algo/optimization specialists, and Systems hackers. I am particularly interested in using program analysis in order to customize and scale inference to huge data sets while retaining the ergonomics and developer friendliness of a general-purpose programming language.

Find Michael Tingley at

Last Year's Tracks