Building Real-Time Applications: Cyclist Crash Detection
summary
A real pipeline processes phone GPS and accelerometer telemetry from a fitness app into Apache Kafka then into a Python microservice that loads a pickled TensorFlow model and writes detections to an output topic consumed over websockets by a map front end. The demo shows rolling 10 second features, checkpointed state aligned with Kafka offsets, horizontal scaling via consumer groups and Kubernetes statefulset, and per‑inference delay around 40 ms while crash events are successfully detected.
AI-written summary · about
chapters (8)
- 0:00Introduction and motivation
- 4:38Real-time use case overview
- 9:41Data collection and preprocessing
- 14:42Model deployment strategies
- 23:13Building the microservice
- 31:40State management and checkpointing
- 34:32Real-time pipeline demo
- 36:46Q&A on model latency
related talks
23mMaking Sense of Streaming Sensor Data: How Uber Detects On-trip Car Crashes
38mSampling and Modelling the World in Real-Time for Mobility
22mBuilding a movement data and analytics platform to de-stress customers & save lives
30mCrash detector
45mLive build: How to harness streaming data in real time
26mOptimal Approaches for Real-Time Machine Learning with Apache Spark on Kubernetes.
26mBuilding A Python Service Using Open Source Components That Connect, Transform & Deliver Data In RT
42mHow to Leverage the Apache Kafka Ecosystem to...