DevOps drives success and value from various use cases
Rapid Proof of Value
Building ML base solutions taking data science outputs e.g. Jupyter notebooks, trained model and AWS ML Services, and through DevOps practices quickly enable business validation of the project hypothesis and showcasing the capabilities of ML.
Productionisation
Productionising ML solutions, ensuring they are highly secure, reliable, performant and cost efficient.
The deployment of ML models and associated components using build tools and Infrastructure as Code to provide consistency and traceability.
Operationalisation
The ongoing automated monitoring and alerting is built into solutions to ensure high availability and that target response times are met.
Providing KPI’s to judge business impact and service levels are understood, measured and achieved
Uniquely integrating tools such as AWS X-Ray to provide complete observability of ML in production to identify issues, bottlenecks or potential optimisations.
Data Science Model Retraining
One the most critical aspects that our DevOps practice leads on is retraining of models and using multiple deployment options (depending on the customer need) when updating them.
Use of Continuous Integration and Continuous Deployment pipelines that take new data, cleans and transforms the data, runs incremental training of ML models, verifies their accuracy and then deploys them into production.