Interview
summary
Machine learning platform work at Coinbase focuses on shipping models to phones for on-device fraud detection and reducing hosted compute costs. The talk traces a live-tweeted project to build a mobile ML pipeline, names tools like Swift for TensorFlow interest and mentions Flink, and reports concrete benefits: model evaluation moves from servers to customers’ phones. Practical lessons include engineering vs research gaps, deployment pitfalls encountered during the prototype, and clear takeaways on where mobile ML can realistically go next.
AI-written summary · about
related talks
43mApplying Machine Learning in the Mobile World: Experiences and Use Cases
09mMachine Learning on the Edge
29mPushing Machine Learning into your app When TensorFlow and MLKit rule the world
25mGetting Started with Machine Learning in your iOS apps
50mBuilding Intelligent Mobile Apps using Machine Learning
39mMachine Learning on Android Demystified
47mMachine Learning on Android Demystified
32mMaking Models Work: How an optimized ML lifecycle can drive better ROI