AI-Ready Microservice Architecture
summary
A working event driven pipeline that uploads images to an object store, triggers a logo detector that emits bounding boxes and confidence scores, pushes synchronous results via a websocket apply service, captures thumbs up/down feedback into a separate feedback broker, and aggregates predictions plus feedback in an analytics DB for retraining. Implemented with Knative Brokers and Channels and Knative services, integrated with KServe and Kubeflow for model lifecycle. Demo runs but the live image demo has reliability issues, illustrating eventual consistency and operational failure modes.
AI-written summary · about
chapters (5)
- 0:00AI-ready microservice intro
- 1:25Logo detection challenges
- 4:57Splitting monoliths into microservices
- 8:30Event-driven architecture demo
- 12:24Key takeaways
related talks
01h 00mEventing with Knative and Cloud Run: From Basics to Advanced
46mAccelerating Machine Learning DevOps with Kubeflow
40mVertex AI: Pipelines for your MLOps workflows
17mAI/ML demo derby
41mAdding Generative AI to Real Time Streaming Pipelines
33mUsing Promptable Image Analysis for Clinical Diagnostics
28mPivotal Cloud Foundry, Google Machine Learning, and Spring
43mAI/ML Data Pipeline Processing with Go Microservices Based...