Antoine Atallah

About Me

My name is Antoine Atallah, and I'm a hands-on computer engineer and data scientist, with a strong passion for large scale systems, robotics and machine learning.

I've always tinkered with things, from taking apart toys to programming computers since I was 8 years old. I then diverted into the world of electrical engineering, and eventually into the world of data science.

Learning new things is very important for me, and have been toying with artificial intelligence and robotics since way before they became a hype ;). I enjoy building things that touch both the computer and the physical world.

I am originally from Montreal, Canada; and have been living in Seattle, USA since 2007.

When not working, or doing projects, I enjoy doing photography and playing music.

Contact Details

antoine at blurbery dot com

Work

Hungryroot

Vice President of Data Science • December 2021 - Present

Food, People and Machine Learning.

Electronic Arts

Group Technical Director • March 2018 - December 2021

Worked as a member of the Frostbite Architecture Team, on core parts of the Frostbite Game Engine for 3+ years.
Worked as a Director on EA Digital Platform, leading a team working on Machine Learning, Cloud and Analytics projects.

Apple

Engineering Manager • June 2015 - March 2018

Lead a team of engineers and designers on the core machine learning team, building platforms and doing research work for the various products at Apple.
We shipped the smarts of the Siri Watchface in collaboration with the amazing Apple Watch folks!!!

Turi (formerly Dato (formerly Graphlab)) (acquired by Apple)

Principal Data Scientist and Software Engineer • June 2015 - March 2018

Work on the core machine learning platform, building "toolkits" to allow people to do actionable things with machine learning:
- Adaptive Machine Learning, online learning on top of any existing machine learning techniques. Extends the longevity of ML models in production, and captures intra-day trends.
- Click Through Rate Prediction, predict the probability of a user taking an action. Model trained in real-time given a feedback signal provided by the target application.
- Churn Prediction, a package allowing people to extract information about which user is likely to stay a customer, and who is likely to leave, provided user-activity data.

Twitter

Staff Software Engineer and Tech Lead • April 2014 - June 2015

Project/Tech lead of a small team working on user modeling.
- Work on user modeling: Understand who users are based on their behavior and profile, at the scale of Twitter (hundreds of millions of users doing billions of things per day!)
- Work on ads prediction: Given an advertising and an audience, compute the probability that a given user of the audience will interact with the advertising.
- Work on ads targeting: Allow advertisers to select which audience they want to target.

Facebook

Senior Software Engineer and Tech Lead • November 2010 - April 2014

Working on large-scale online machine learning systems that can grow at Facebook Scale (1 Billion monthly users hitting the site hundreds of millions of times a day).
- Work on recommender system infrastructure for suggestions (group suggestions, page suggestions, etc...)
- Work on groups recommendation (Facebook groups you should join)
- Work on large-scale data analysis of user behavioral data
- Ads ranking using large scale real-time machine learning model
- Sponsored Results, user behavior prediction on free-text searches
- Data analysis using Hadoop, Hive and Giraph

Tokamak Investment Group

Co-Founder • May 2009 - September 2012

Co-founded (along with two other partners) a small Automated Trading company, investing private funds and performing day-trading activities. The system would use various machine learning techniques to determine the best stocks to buy and sell, and automatically perform the appropriate book-keeping. This would guarantee compliance with the SEC rules, while optimizing the trading strategies.
The three of us were involved in writing code, deployment, statistical modeling of the market, implementation of machine learning algorithms for real-time/streaming applications, including real-time training and prediction.
The system was written in C# and ran on a variety of redundant systems on privately owned hardware.

Microsoft

Senior Software Engineer • July 2006 - November 2010

Working on experimental technologies at Bing:
- Self-tuning online machine learning algorithms
- Data mining and data analysis of user behavioral data
- Work on artificial vision projects - Recognizing visually similar images using neural networks
- Work on Microsoft Office - Office Online and Excel
- Multiple patents in the fields of: Artificial vision technologies, Machine learning

Mechtronix

Computer Engineer • March 2005 - July 2006

Develop hardware device drivers, and work on the core application of the flight simulators. Help with debugging hardware and software issues in the simulators as well as build prototypes using new technologies.
Work under the QNX real time operating system.

Entreprise YWait inc

Director - System Architect • June 2002 - February 2005

Lead and work with a small team developing systems in logistics for the transportation industry. Problem centered around an online (real-time), more complex version of the travelling salesman problem, solved using different heuristics and advanced algorithms.

Education

Concordia University

Bachelor of Engineering • May 2005

Undergrad in electrical and computer engineering, specializing in hardware design. Did a few masters level classes in artificial intelligence, and on-the-side learned about machine learning.

Skills

  • Talking to people incessantly (team lead)
  • Agonizing over small details so nobody else has to (systems architect)
  • Making machines talk to one another (distributed systems)
  • Making machines learn (machine learning)
  • Getting my hands dirty (software and hardware development)
  • Make big things (scalable systems / service oriented architecture)
  • Multilingual (C++, Python, C#, Scala, PHP, Objective-C, Swift, way too many)