FeatureByte launched by Datarobot veterans to advance AI feature engineering

Artificial intelligence (AI) holds great promise for companies to help optimize processes and improve operational efficiency. However, the challenge for many is getting data in the right form and with the right processes to actually benefit from AI.

That’s the challenge that FeatureByte’s two co-founders, Razi Raziuddin and Xavier Conort, noted time and again while working at the enterprise AI platform provider. computer robot. Raziuddin worked for more than five years computer robot including a stint as senior VP of AI services, while Conort was the chief data scientist at Datarobot for over six years.

“One of the challenges we’ve seen is that AI isn’t just about building models, which is the focus not just of Datarobot, but pretty much all of AI and ML. [machine learning] tool room,” Raziuddin told VentureBeat. “The main challenge that still exists, and we call it the weakest link in AI development, is only the management, preparation and deployment of data into production.”

Borrowing data preparation from data analytics to improve AI development

Raziuddin explained that feature engineering is a combination of several activities designed to help optimize, organize, and monitor data so that it can be used effectively to build features for an AI model. Feature engineering involves data preparation and ensuring that data is in the correct format and structure to be used for machine learning.

In the world of data analytics, the data preparation process is not a new discipline; there are ETL (Extract, Transform, and Load) tools that can extract data from an operational system and then convert it into a data warehouse where the analysis is performed. That same approach was not available for AI workloads, according to Raziuddin. He said data preparation for AI requires a purposeful approach to help automate an ML pipeline.

To do really good feature engineering and feature management, Raziuddin said it takes a combination of several critical skills. The first is data science, with the ability to understand the structure and format of data. The second critical skill is understanding the domain in which the data is collected. Different data domains and use cases in the industry will have different concerns about data preparation, such as data collected for a healthcare implementation will be very different from that for a retail store.

With a deep understanding of the data, it is possible to build features in AI that are optimized to make the best use of the data.

How FeatureByte wants to automate feature engineering for AI

Getting data into the right shape for AI often required a data engineering team alongside one or more data scientists.

What FeatureByte wants to do is help solve that pain point and provide a streamlined process for having data pipelines available that data scientists can use to build features for their AI models. Raziuddin said his company is really all about taking friction out of the process and ensuring that data scientists can do as much as possible within one tool, without relying on a data engineering team.

FeatureByte’s technology is still under development, although the company has some clear goals for what it should be able to do. Today, it announced that it has raised $5.7 million in an initial funding round. Raziuddin said the platform will use the funding to embed domain knowledge and data engineering expertise to accelerate the feature engineering process.

FeatureByte’s platform will be cloud-based and will be able to leverage existing data sources, including cloud data warehouses and data lake technologies such as Snowflake and Databricks.

“As the number of AI models grows, the number of data sources available to build these models is increasing at a faster rate than most teams can handle,” said Raziuddin. “So unless there’s tooling and unless that process is automated and streamlined, companies won’t be able to keep up.”

The seed funding was led by Glasswing Ventures and Tola Capital.