The bumpy road of bringing a machine learning model from development to production - Part 2 Infrastructure, testing, productionalization and monitoring - FOSSASIA Summit 2023

FOSSASIA Summit 2023 Thursday, 13 April, 2023 08:30 (Asia/Singapore) To Saturday, 15 April, 2023 18:30 (Asia/Singapore)

Info Featured Schedule Speakers Call for SpeakersCfS

The bumpy road of bringing a machine learning model from development to production - Part 2 Infrastructure, testing, productionalization and monitoringWe will report on our journey from developing and deploying a machine learning model in the search space for the biggest C2C e-commerce marketplace in Japan. This talk will focus on the MLOps side of this journey (with a separate one focusing on the machine learning part). In particular, we will discuss the following topics:<ul><li>Using ML in a “traditional” search infrastructure</li></ul><ul><li>Architecture overview</li></ul><ul><ul><li>Data engineering</li></ul></ul><ul><ul><li>ETLs</li></ul></ul><ul><ul><li>Training pipelines</li></ul></ul><ul><ul><li>Serving in production</li></ul></ul><ul><li>Strict latency requirements</li></ul><ul><ul><li>Timeouts and failsafes</li></ul></ul><ul><ul><li>Caching</li></ul></ul><ul><li>“Productionalization”</li></ul><ul><ul><li>Logging</li></ul></ul><ul><ul><li>Monitoring</li></ul></ul><ul><ul><li>Metrics</li></ul></ul><ul><li>A/B testing with ML</li><li>Moving from “homegrown” model server to Seldon for model serving</li><li>Our learnings and key takeaways</li></ul>During all the steps, we have used open source tools for the development and deployment.

0 Comments

You must log in to leave a comment.