AI models are increasingly being deployed into medical devices and systems. Companies are very careful to ensure they're safe, but is the development fast and efficient?

Want faster model development? Get the "2019 State of Development and Operations of AI" report now.

Get the report

Companies in the healthcare space have a particular responsibility to ensure that the technology they deploy is safe and well-tested. The current process for doing safe machine learning development, however, is far from optimal — it involves extremely meticulous manual logging of every dataset and step that went into training a model, and exactly how it was achieved.

What if there was a way for ML model development to be both fast and safe?

Key AI challenges

  • Innovators in healthcare have realized that they need to develop model management systems which track the evolution of data, code and models together in a consistent way.
  • Creating these systems in-house however is a multi-year process, and such systems-level software is not the core competency of these companies.

  • Tracking labels in a SQL database and data in cloud object storage creates barriers and opacity between data engineering and model development teams.
How can Dotscience help?

Dotscience takes tried and tested DevOps techniques enabling both fast and safe development, and applies them to the increasingly complex world of data and models. For example, Dotscience can track data and labels from multiple sources, and maintain a provenance trail from models that are deployed into the field all the way back through all the steps of model development and data engineering. Furthermore, collaboration is accelerated as any ML or data engineer can take another's run and fork it, then propose the changes back asynchronously, accelerating the exploration of different approaches to further increase model accuracy and efficacy.