IoT has enabled companies to access data from billion of devices from all over the world but the full exploitation of this data has not yet been achieved as AI adoption still lags. This is mainly due to the fact that 90% of Machine Learning models do not reach the production stage because of a rapid decline of model performance in the industrialization phase. It is therefore necessary to provide solutions which can facilitate the transition from development to production. This is where FIWARE and MLOps come in to play to provide a full platform that can monitoring your data and models through different mechanisms and strategies to limit or prevent performance degradation.

Speaker: Yannick Lecroart (MLOps / Data Scientist, ATOS)

Chapter: Operations
Difficulty: 4
Audience: Data Scientists, Operations Engineers

0:00 Introduction
1:00 What is MLOps?
6:13 Tooling
13:52 Model & Data Monitoring
19:06 Continuous Integration Deployment and Training
22:12 Further Experimentation and Learning
31:17 Summary

свернуть