The commoditization of Machine Learning training tools and proliferation of DevOps tools has provided a variety of paths for transitioning experimental code to production systems, but performing this transition responsibly requires skill sets in Data Science, Machine Learning, Software Engineering, and DevOps among others. As a result of the pace of innovation and lack of standardization in the required fields, a "Valley of Death" has emerged between Machine Learning capability development and production deployment.
The OMI has been founded to provide a spec that can reliably guide organizations as they transition Machine Learning models into production using a simple yet powerful interface that:
- standardizes certain security best practices
- supports the widest range of machine learning and artificial intelligence training tools and frameworks
- enables re-use of existing integration code to add new or updated models to production environments by sharing a common API across all model containers