Large Scale Distributed AI on the JVM
summary
Shows a working pipeline: 125 Raspberry Pis cluster for parallel Java workloads vs a 512-core Jetson TX2/Xavier for GPU inference, plus a benchmarking demo comparing CPU vs GPU on time‑series. Uses Kura/OSGi gateways, MQTT, LoRa radios and edge anomaly detection handling 6,000 messages/sec and 1 GB acoustic captures; deploys TensorFlow models via TensorFlow Lite or IBM PowerAI in minutes and demonstrates object detection APIs and on‑device vs cloud tradeoffs.
AI-written summary · about
chapters (40)
- 0:00Introduction
- 1:48Journey in Java
- 4:51Sunspot
- 5:42PhD
- 6:57Current project
- 9:25What is AI
- 10:14Machine Learning Deep Learning
- 10:59Deep Learning
- 11:37Machine Learning
- 13:20Olympic Park
- 15:13Communication
- 17:08Devices
- 17:25Radios
- 18:37Sensors
- 19:54Gateway
- 22:19ML United
- 24:38Anomaly Detection
- 25:28IoT Signals
- 26:17Latency
- 27:06Gate Phase
- 28:52OSGI
- 32:03Oracle
- 33:00Visual Recognition
- 35:45Image Recognition
- 36:07Tensorflow Lite
- 36:40CPU vs GPU
- 37:02Nvidia Jetson TX
- 37:53Jetson Xavier
- 38:38Tensorflow
- 39:55APIs
- 40:26Model Asset Exchange
- 41:42Netbeans
- 47:52IET Platform
- 49:15IET in IBM
- 49:29Analytics
- 50:42Visual Integration
- 51:18Transportation Utilities
- 51:46IoT Services
- 52:00IoT Security
- 53:16Hybrid
related talks
26mBuilding a Cloud Native Edge Platform for Accelerating AI Inference
38mAI in Kubernetes: How to Get Started?
16mUsing AI and Computer Technologies to Enhance IT Infrastructure
59mML and the IoT: Living on the Edge
20mOperating Frontier Intelligence at Scale
14mAccelerating AI inference workloads
27mHarnessing Kubernetes for Scalable AI/ML Workloads
36mApplied AI on the edge