Description:

As part of the ML Lifecycle, Data Scientists and Engineers often face the issue of creating and re-creating features, checking data and creating ML models. In order to help streamline this process, Feature Stores are an important part of the MLOps pipeline, which not only helps in keeping store of the features being used in ML models, but also helps streamline the processes for deployment and monitoring of models.

Through this workshop, participants will be able to understand how to create and use feature stores, optimizing workflows for creation and monitoring of ML models and some tips for future enhancements to the feature stores.

This workshop aims to cover the following---

• Introduction to Feature Stores
• Implementation of Feature Store
• Feature Engineering with Feature Store
• How to integrate feature store into ML Pipelines?
• Monitoring and Governance steps for Feature Stores

In order to allow for accessibility and, This workshop will not use any specific cloud platforms like AWS or Azure or use any cloud specific products like AWS Sagemaker. The workshop will instead aim to create a feature store using traditional Python packages like PySpark, pandas etc so that participants can understand the traditional steps for creation and management of feature stores and can transfer that knowledge into the respective cloud platforms.

Prerequisites:

Participants should preferrably have good knowledge of PySpark, pandas and other traditional ML libraries like XGBoost, sklearn etc. A deeper understanding of pyspark would not be necessary for this workshop.

Speaker Info:

Sourav Singh works as a Data Engineer in a large scale bank, where he works on managing and deployment of new feature stores and helping Data Scientists develop new models at a faster rate. He has been an active contributor to Open Source and had been a speaker for PythonPune, PyCon India and also worked as a coach for Django Girls Pune and Bengaluru chapter. Apart from this, he also likes to workout and tinker with Gunpla models.