“AI is the new electricity” says Andrew Ng, one of the pioneers of AI and online learning. His remark may be an overstatement, but there’s no denying the impact AI has had, is having, and will continue to have on our lives, at both personal and professional levels. One of the core instruments of AI is Machine Learning, and by attending this workshop you will get the chance to learn about ML and understand how to use it effectively.
During this workshop, you will learn about the types of problems that can be solved by Machine Learning, how to assess a Machine Learning problem, about training and evaluating a predictive model, and deploying that model in production.
After attending, you will be better equipped for applying standard, proven software engineering techniques such as version management, automated testing, and continuous integration and deployment to machine learning projects.
This is a technical training targeting software developers with basic machine learning and Python knowledge, yet it can also be updated to accommodate software developers with little to no machine learning experience. The recommended number of participants is 10 to 15.
Included below are the main topics covered in this workshop. The recommended duration for the course is 3 days however if needed, the workshop can be condensed into 2 days.
- Introduction to machine learning in the Azure cloud
Discuss the various flavors of machine learning in the Azure cloud.
- Development environment
Discuss and showcase popular open-source development tools, such as Jupyter Notebooks and Visual Studio Code.
- Identifying and solving machine learning scenarios
Work alongside the attendees to solve basic regression and classification machine learning scenarios, using several approaches such as via the Azure Machine Learning Visual Interface, using Azure Automated ML, and also using open-source libraries such as scikit-learn.
- The machine learning workflow
Discuss the components of the machine learning workflow such as asking the right question, gathering data, choosing a feature set and an algorithm, selecting an appropriate metric and using that metric to evaluate the performance of the algorithm. Work with the attendees to identify the right metrics and evaluate the models trained previously.
- Tracking machine learning experiments
Discuss the importance of tracking model performance while iterating across multiple experiments, working alongside the attendees to track several versions of their models using open-source tools such as MLFlow, and more established ones such as Azure Machine Learning Experiments.
- Deploying machine learning models in the cloud
Work with the attendees to deploy multiple models in the Azure cloud as secure REST services, using several approaches. Integrate with the deployed models, and discuss the advantages and disadvantages of each approach.
- Continuous deployments for machine learning
Discuss available options for continuous deployments, practice deploying models using Azure DevOps pipelines and Azure Machine Learning Workspace pipelines. Build pipelines that continually retrain and update model deployments.
- Automated Machine Learning in depth
Train models using automated machine learning for regression/classification/time series forecasts, see advanced configuration options and learn how to analyze and update automatically trained machine learning models.
- Model explainability
Analyze and understand how trained models interpret their inputs, determining whether hidden bias exists, which features the model deems important and which are ignored.
About the trainer
Vlad Iliescu is a Partner and Head of AI at Strongbytes, a razor-sharp company with a strong focus on building software products around well-operationalized machine learning models. He organizes NDR, an annual machine learning conference in two of Romania’s largest cities – Iasi and Bucharest.
He is fortunate to be a Microsoft Most Valuable Professional on AI, a group of technology experts recognized by Microsoft for their passion about sharing their knowledge with the community, of which worldwide, there are less than 100 MVPs on AI.