Operationalising Data Science on Cloud Foundry
summary
Argues Cloud Foundry is practical for running and scaling models and shows a live demo: an accelerometer app built as three microservices (training, scoring, front end) that trains a random forests classifier, stores models, and serves real-time predictions. Compares patterns: scoring-as-a-service, CF-powered learning, and online in-stream updates. Covers tooling: Python buildpack, Spring Cloud Data Flow, Redis/marketplace services, PMML/PredictionIO, and gaps around model versioning, replayability, large-batch training and the need for a coherent model service and isolation segments for heavy compute.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 0:55Who am I
- 1:38What are we talking about
- 2:19The problem
- 3:09Impact
- 4:35Running predictive models
- 6:30Scoring as a service
- 7:47CF powered learning
- 9:25Online learning
- 10:25Marketplace services
- 11:51Demo
- 14:09Scoring Phase
- 14:48Next steps
- 16:50Model service
- 18:43Whats next
- 20:28Questions
related talks
32mTransforming the Retail Experience with Cloud Foundry and Machine Learning - Michael Goddard
37mJava Microservices at Scale on Cloud Foundry, Today and Tomorrow
12mDemo with Emamurho Ugherughe
17mProductionizing Data Science Workloads
31mCloud Foundry and Microservices: A Mutualistic Symbiotic Relationship
49mDDD Strategic Design with Spring Boot
35mDoing Data Science on Cloud
33mHybrid Cloud in Finance