LinkedIn, a business networking site owned by Microsoft Corp., is giving another project to the open source community. It announced today that it is giving the AI & Data Foundation of the Linux Foundation authority of the Feathr feature store for machine learning applications. An instrument called Feathr was created by LinkedIn and launched in 2017. It is intended to simplify, accelerate, and scale up feature serving in machine learning, particularly for real-time AI applications. According to a blog post by LinkedIn engineers Hangfei Lin and Jinghui Mo, the company’s AI teams use Feathr to store, transform, serve, and distribute features with high throughput and low latency.
Feathr, in particular, acts as an abstraction layer between the raw data and machine learning models, assisting in standardising and making workflows and applications for machine learning workflows and apps for feature definition, transformation, serving, storage, and access easier. When Feathr connects to numerous databases, developers can concentrate more on feature engineering while letting Feathr to handle data serialisation types. Additionally, it offers options for managing credentials and optimising performance.
Feathr enables authors of machine learning algorithms to create features only once and use them across a variety of contexts, including model training and model serving. Additionally, they have the ability to connect to multiple offline data sources, like data lakes and data warehouses, and convert the data contained within into machine learning characteristics. Since its debut, Feathr has grown to power numerous AI applications at LinkedIn, where it is employed to manage a large number of features. Lin and Mo claim that it has enabled teams to experiment with and implement new features more quickly—from weeks to days—while outperforming the proprietary feature processing pipelines it replaced by up to 50%.
In April 2022, Feathr underwent its initial Apache-2 licence open sourcing. The business also revealed native Feathr support and integration on Microsoft Azure at the same time. Since then, Feathr, according to Lin and Mo, has experienced a surge in popularity among machine learning operations professionals and is now utilised by businesses of all sizes and across a variety of sectors. Additionally, those individuals are not only using the programme but also actively participating in its development. “It’s clear that many others experience the same pain points that Feathr aims to address,” Lin and Mo said.
It is intended that by giving Feathr to the LF AI & Data, it will be able to expand and evolve more quickly, boosting its visibility, user base, and contributor base. The main Feathr development team anticipates increased chances to work with organisations and initiatives that are utilising the programme. One goal is to enhance support for online retailers, perhaps through integration with Milvus and JanusGraph. Additionally, supporters of Feathr aim to incorporate the OpenLineage open data lineage standard.
“We’re excited to welcome Feathr to LF AI & Data and for it to be part of our technical project portfolio with more than 16,000 developers,” said Dr. Ibrahim Haddad, executive director of LF AI & Data. “We aim to support Feathr to expand its user base, grow its community of developers, become a leader within its own category, and enable collaboration and integration opportunities with other projects.”
However, not everyone was won over by LinkedIn’s show of generosity. Although open-sourcing Feathr is a commendable initiative, the market is already quite crowded, according to Andy Thurai, vice president and principal analyst at Constellation Research Inc.
