Activeloop Raises $5M Seed for AI Streaming Database

Activeloop Secures $5 Million Seed Funding for AI-Focused Database

Activeloop, a company originating from the Y Combinator summer 2018 program, is developing a database tailored for artificial intelligence applications dealing with media. The company has recently announced a $5 million seed investment round. This funding was spearheaded by 468 Capital and CM Ventures, with further participation from Tribe Capital, Shasta Ventures, and a group of technology industry investors.

Origins and Development

Davit Buniatyan, the founder and CEO, explains that the company’s inception stemmed from his research conducted at Princeton University. He identified a critical need for a streaming database capable of handling unstructured data – such as images and video – specifically for AI workloads. Concurrently, the company is initiating the alpha release of its first commercial product.

Core Functionality: A Database Designed for AI

“We are introducing a database solution designed specifically for AI,” Buniatyan stated. “This includes a storage layer engineered to efficiently store data and then stream it to machine learning applications for tasks like computer vision, audio processing, and natural language processing (NLP).”

How the Technology Works

The technology manifests as an open-source API. This API empowers users to transform data – including video and audio files – into a mathematical format understandable by machines. Furthermore, the API facilitates version control of the data and its storage within repositories like Amazon S3.

Streaming Data for Efficiency

By employing data streaming, data scientists and developers circumvent the necessity of downloading entire datasets to their local machines. Instead, data is transmitted for model utilization in a manner analogous to how Netflix streams movies, avoiding complete downloads.

Community and Adoption

Currently, the open-source project boasts 55 contributors and a community of 700 members. The software has been downloaded over 300,000 times to date.

The Inspiration Behind the Streaming Approach

The concept of streaming data arose during Buniatyan’s work with substantial datasets at Princeton’s neuroscience lab. The sheer size of the files rendered traditional downloading impractical, leading to the development of this streaming methodology, which ultimately formed the foundation of Activeloop.

Team Growth and Diversity

The company presently employs 15 individuals and intends to recruit approximately six experienced engineers. Buniatyan emphasizes a commitment to seeking diverse talent globally. He highlights that the company’s VP of Engineering is a woman, and recounts a recent job offer extended to an engineer from Ethiopia, though ultimately declined, as an illustration of this commitment.

Focus on Mission and Impact

“The origin of a person, or potential biases related to their birthplace, are not primary considerations,” he explained. “What truly matters is our shared mission – building a company that makes a significant impact. We value team members dedicated to realizing this vision.”

Commercial Offering and Future Plans

While the API remains open source, the company’s commercial product, now in alpha, will introduce additional features. These include the capability to execute SQL queries on the streaming data, a functionality not available in the open-source version. Activeloop is currently pre-revenue, with plans to launch the full commercial version in the first quarter of the following year.