Course Description: Blent’s Machine Learning Engineer bootcamp is about a highly sought job in Data Science. The training, as the two previous ones, is a part time online training and is supervised by highly qualified mentors. The in-depth knowledge of a Machine Learning Engineer enables him or her to handle large-scale projects. The ML Engineer knows how to deploy Machine Learning models, industrialize projects and manage highly scalable infrastructures.
Blent’s Machine Learning Engineer bootcamp is about a highly sought job in Data Science. It is a part time course over 3 months, perfectly adapted for workers, entrepreneurs and students. It iis supervised by highly qualified mentors. The in-depth knowledge of a Machine Learning Engineer enables him or her to handle large-scale projects. The ML Engineer knows how to deploy Machine Learning models, industrialize projects and manage highly scalable infrastructures.
Program :
As during this training, a multitude of different tools are needed, it begins with a preparation bootcamp. This latter is meant to pave the group to many advanced concepts of the Python language such as serialisation tools or environments. Acting on different systems, the future Machine Learning engineer needs to know the Bash and UNIX scripting’s basics, as well as project collaboration with Git. Finally, as Cloud Computing has its own vocabulary, this part is detailed in the preparation session too.
After the preparatory bootcamp, come other phases to be known: ML pipeline, model deployment and MLOps.
The ML pipeline consists of unifying the data collection and processing steps with modeling steps.
To do this, we will see how Bayesian optimization methods will be used in the future useful for automating modeling and determining the most efficient model. Since the data is often voluminous, Apache Spark is used for manipulating data with parallel computing and storing it in a NoSQL database with Google BigQuery.
Then the phase Model deployment is the core business of the ML Engineer: putting into production models on servers in the Cloud. For this phase many tools will be used, Kedro (for the architecture of the ML pipeline) and MLFlow (for models filing and tracking) and API REST / Docker.
The third phase is the MLOps one. It is a discipline of automating the deployment of Machine Learning through pipelines. Mixture of DevOps and Data Science, the MLOps is an indispensable skill for the efficient production of models. This phase of the training will start with self-managed serverless deployment using triggers of CI/CD pipelines then the deployment of the environment of production on Kubernetes. Then, in order to be updated regularly with more recent data, Airflow enables model training and deployments to be performed without intervention manual.
Finally, basic principles of safety and tools of monitoring are going to be taught, which are crucial points in the production environment.
Resources : Apache Spark, Google BigQuery, PySpark SQL, BigQuery, LightGBM, MLflow, Docker, Cloud Build, Stackdriver, Grafana