Saving Human Lives with the IoT
summary
Project targets predicting wet or slippery road conditions to reduce part of the 1.2 million annual road deaths by combining car telemetry and weather feeds into a Lambda-style pipeline on Pivotal Cloud Foundry. Ingest uses Spring XD to S3, PySpark on EMR for feature prep, GPU training with Keras/TensorFlow, model cached in Redis and exposed as a predictive API stored in Redis for microservices and a dashboard. Imbalanced labels (~80/20 no-rain/rain) push precision/recall metrics; GPU training is 10x faster and a working prototype ships in seven weeks.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 3:34Technology
- 9:35Data Science
- 12:33Neural Networks
- 14:56Recurrent Networks
- 16:10Learnings
- 19:25Process
- 20:20API First
- 21:46API First Idea
- 22:21Waterfall Culture
- 25:10Key takeaways
related talks
01h 15mIoT Realized - The Connected Car v2
01h 09mIoT Realized - The Connected Car
22mBuilding a movement data and analytics platform to de-stress customers & save lives
52mMachine Learning in IoT Applications
20mAutomating Data Science for the IOT
37mFrom Sensors to Insights at the Edge: Applying AI to IoT Data
47mMore than rubber on the road: Tires in an IoT world
29mML-Powered IoT: From Warehouse Data to Prod Intelligence in Real Time