Data Minds #3— Jonathan Morra
In this episode, I interview Jonathan Morra (VP of Data Science at ZEFR).
You will learn:
- Jonathan Morra’s path to data science
- Jonathan’s data science career advice
- how matching works at eHarmony: Compatibility, Affinity and Distribution
- the main tools used by data scientists at eHarmony
- awesome algorithms and frameworks you may not have heard of
- how data teams and product teams can work together more effectively
- opportunities for data science tool creation
- the importance of community, evangelizing and storytelling in data science
*Note: at the time of this recording, Jonathan Morra was the Director of Data Science at eHarmony. Currently, he is VP of Data Science at Zefyr.
Meet Jonathan Morra
Jon Morra obtained his BSE from Johns Hopkins in Biomedical Engineering (BME). Jon also got his MSE and PhD in BME from UCLA. After finishing his degrees he worked for a small computed radiography company in Torrance where he wrote radiological software (PACS). After a year he got the itch to start his own company and co founded Medical Vision Systems where he developed a machine learning based approach for image segmentation within pre op radiation oncology patients. Through MVS he was a co author on a winning NIH SBIR grant. After MVS Jon co founded The Oncode Group to create dosimetry software to help emergency workers.
After The Oncode Group Jon took a role at eHarmony as a Senior Software Engineer in Machine Learning.
At eHarmony he developed machine learning methods to help with image understanding, matching, and fraud amongst other things. He has open sourced some of his own software and is a contributor to others including a model representation framework, Aloha, and a large scale linear learner, Vowpal Wabbit. Jon rose through the ranks at eHarmony eventually becoming the Director of Data Science where he oversaw all machine learning efforts within the company. Recently Jon took a role at ZEFR as the Vice President of Data Science. In that role he oversees all machine learning and data science efforts to produce data driven models that help ZEFR deliver better content to its customers.
Remember: subscribe to Data Minds!
Get notified of new interviews with top data minds from around the world; subscribe to Data Minds.
Selected Links from the Episode
- Vowpal Wabbit — “probably the fastest linear learner in the world… we use it a lot at eHarmony.”
- Spotz — a hyperparameter optimization framework written in Scala designed to exploit Apache Spark to perform its distributed computation
- Face Parts Service — A RESTful web service for detecting faces, pose, and fiducial markers from an image
- The Big Short
- Ensemble Learning
- AdaBoost
- Voxels
- SBIR — Small Business Innovative Research grant
- Map Reduce
- The Encode Group
- Aloha
- Scala
- Functional Programming
- Hyperparameter optimization
- hyperopt
- h2o.ai
- Spark
- Multi-armed Bandit
- Immutability
- ICML
- KDD
- NIPS
- Szilard Pafka’s machine learning benchmark comparisons
People Mentioned
Show Notes
- Jonathan’s Path to Data Science:
Biomedical engineering undergrad at John Hopkins => Biomedical Engineering PHD => understanding changes in brain => Tracing Voxels by hand to Automation with Machine Learning => Dev Work => Founded Medical Vision Systems (2011) => Web Dev in Rails => Vaclav speaks at ML Meetup in LA => Senior Software Engineer in ML at eHarmony => Director of Data Science at eHarmony - Open Source Projects… Vowpal Wabbit
- Founder Lessons (wearing many hats). The importance of interpersonal skills.
- Vaclav speaks at an ML Meetup in LA => Jonathan joins eHarmony as Senior Software Engineer in ML at eHarmony
- Mentors and support group during the journey
- Befriending Ryan Deak. Scala, Aloha and falling in love with functional programming.
- How Matching at eHarmony Works: Compatibility, Affinity and Distribution
- Day in the Life of Running the DS Team at eHarmony
- The Modeling Team, Evangelizing Work & Storytelling, Working with Product Teams
- Advice to Product Teams: The Importance of Immutability
- Jonathan’s Two Recommended Data Science Paths
- Main tools at eHarmony:
- Scala group inside a Java shop
- Apache Spark
- Vowpal Wabbit
- H2o.ai - DS Pet Peeves — the definition of data science
- Importance of Different Backgrounds
- Future of Deep Learning
- Opportunities for tool creation
- Autonomous Machine Learning and Human Understandable Machine Learning
- Recap of Tools & Libraries
- Rapid Fire Questions
Subscribe to Data Minds
Get notified of new interviews with top data minds from around the world, subscribe to Data Minds.
