A researcher would email a serialized model and instructions to an application team. Registries, versioned artifacts and deployment pipelines replaced the informal handoff.
Manual model handoff no longer exists as a living trade. Here is what erased it, and which profession took on the work.
1950s–2005
Models as research artifacts
Machine learning lived largely in research, specialist products and statistics. Operational concerns existed but were not usually a distinct career.
What else was happening then
1998
Google's PageRank highlights data-driven systems
Web-scale ranking shows how models and data infrastructure can become central product infrastructure.
2006
Cloud computing becomes mainstream
Amazon Web Services makes elastic compute and storage available as operating services rather than owned hardware.
2012
ImageNet breakthrough accelerates deep learning
AlexNet's success on ImageNet makes GPU-based training central to many machine-learning programs.
2015
Hidden technical debt is named
Google researchers describe why ML systems accumulate dependencies and maintenance burdens beyond their model code.
Where that work lives now
Builds the systems that train, deploy, monitor and govern machine-learning models in production.
📦 MLOps Engineer →