Companies have invested a lot of effort and money in data-driven strategies. However, many also testify to the difficulty of deriving the full ROI from these initiatives. Data Science Use Cases should not be any more pilots, they need to be though as industrialized products from the beginning. As such, they must be managed as business and technical assets that need to be developed, deployed, monitored and improved over time. As software artefacts, they can largely benefit from software industry practices such as DevOps.
MLOps has adapted DevOps practices to the development and operation of data science use cases, more specifically use cases embedding machine learning models.These machine learning models, as part of the use case deployed, need to be monitored, evaluated, retrained and certainly improved over time.
It becomes critical to consider the whole life cycle of data science use cases with a robust methodology and set of practices. These will ensure an efficient design, delivery and continuous improvement of machine learning based use cases, to get most of the ROI.
In this webinar, we introduce:
- Key notions of MLOps practices
- Highlight classical difficulties & roadblocks of data science use case life cycle.
We present MLOps practices, illustrate these elements through actual use cases, and how Dataiku DSS can support such MLOps practices.
Data and R&D Associate
* Required fields
Subscription : https://www.brighttalk.com/webcast/17108/469921