Exercise Analysis

on YouTube ↗44m

" summary

Muvr builds a full pipeline: wearables (Apple Watch, Pebble) stream accelerometer and heart data through mobile to an Akka-backed cluster and Spark training. Code bases span C/C++ and Objective‑C++ on devices, Swift on mobile, Scala and Spark for server training. Early deployment fails: 1–1.5s latency and dropped 3G kills UX, so computation moves to on-device classification with multilayer perceptron models, training cycle drops from days to ~2 minutes and evaluation reaches error rates around 0.7%.

AI-written summary · about

" chapters (7)

1. 0:00Start
2. 6:37Example of Accelerometer Trace
3. 7:56Sensor Data
4. 13:48Vehicle Cluster
5. 16:20User Profile
6. 38:48Classification Results
7. 41:50User Interface

" related talks

• 52mMuvrJan Machacek · Lambda Days 2016
• 42mSo, You Want to "Do Big Data?"Jan Machacek · Scala Days Berlin 2016
• 17mAI workflow for large scale deployment of far-edge ML devicesKabir Manghnani, Mark Stubbs · Arm AIoT Dev Summit 2019
• 10mMachine Learning for mobileKemal El Moujahid · Google's Machine Learning Virtual Community Day 2021
• 26mOptimal Approaches for Real-Time Machine Learning with Apache Spark on Kubernetes.Hichem Kenniche · Community Over Code EU 2024
• 06mINTERVIEWAnubhav Shrivastava · Machine Learning Developers Summit 2019
• 23mWhy MLXPrince Canuma · AI Engineer Europe 2026
• 45mAutomated data analyzer using Machine LearningEiti Kimura · DevCamp 2019