Topic

MLOps and production machine learning

Getting a model into production and keeping it there is a different problem from building it. The work here covers that problem from three sides: modernising a data science function at a bank toward a governed path from experiment to production on Databricks, running optimisation and streaming systems for a national telecoms network, and Ubunye Engine, an open source framework whose model registry, lineage and run anywhere guarantee are tested rather than promised.

3 work · 1 writing · Wikidata Q60753505