Prepare to manage the “evolution” of AI-based solutions with “DataOps”
Conclusion: The development of AI-based solutions is heavily dependent on various types of data input in the form of either:
- Large data sets used to conduct experiments to develop models and algorithms for predictive analytics, optimisation and decision recommendations; or
- Enriched and tagged corpuses of images, audio, video and unstructured text used to train neural networks using deep learning techniques.
While at first the data management needs of AI-based solution development might leverage both data scientists and their existing business intelligence platforms to exploit these types of data, the actual lifecycle management needs of AI developers will expand quickly beyond the boundary of the traditional enterprise data warehouse.
Therefore, like the source code and configuration data underpinning transactional business applications, the raw data and algorithms of AI solutions must be managed by evolving DevOps practices towards a comprehensive “DataOps” model.