From 6f99f596f0cc70538e92b89fab01347da730dbce Mon Sep 17 00:00:00 2001 From: Florian <83873752+florianim@users.noreply.github.com> Date: Mon, 16 Sep 2024 01:02:27 +0200 Subject: [PATCH] week 4: presentation proposal (#2482) Co-authored-by: Florian Immig --- .../presentation/week4/prerna-immig/README.md | 27 +++++++++++++++++++ 1 file changed, 27 insertions(+) create mode 100644 contributions/presentation/week4/prerna-immig/README.md diff --git a/contributions/presentation/week4/prerna-immig/README.md b/contributions/presentation/week4/prerna-immig/README.md new file mode 100644 index 0000000000..e8f4c4f830 --- /dev/null +++ b/contributions/presentation/week4/prerna-immig/README.md @@ -0,0 +1,27 @@ +# Assignment Proposal + + +## Title + +What is a Feature Store in ML? + +## Names and KTH ID + + - Prerna Gupta (prerna@kth.se) + - Florian Jerome Immig (immig@kth.se) + +## Deadline + +- Week 4 + +## Category + +- Presentation + +## Description + +We want to do a presentation on the concept of the Feature Store. We will explain the significance of using a feature store when dealing with MLOps workflows and present its advantages. Specifically we will talk about how it enhances real-time (online) applications. In our presentation we use Feast as an example for an open-source feature store and provide according code-snippets. + +**Relevance** + +When not having a feature store, ML feature reusability is limited and ML engineers spent a significant amount of time on feature engineering. The concept of a feature store acts as a hub of callaboration and ensures consistency accross training and serving avoiding training-serving skew. Furthermore, features can be reused accross teams leading to additional time savings. Feature stores allow for a consistent and fast delivery of feature values in online, offline and training scenarios. \ No newline at end of file