Apple has acquired Datakalab, an AI startup that specializes in on-device AI, another significant insight into the iPhone maker’s approach to AI.

According to MacRumors, Datakalab’s technology revolves around embedded AI and algorithm compression. The company was founded in 2016 and specializes in “low-power, high-efficiency deep learning algorithms,” exactly the type of technology needed to deploy AI on a phone or tablet. The company’s algorithms also involved advanced facial recognition, an ideal paring with Apples’ Face ID.

The deal evidently happened in mid-December of last year, but was only recently discovered by French publication Challenges, thanks to an EU regulatory filing. No financial details of the deal were disclosed.

Interestingly, according to RTTNews, the company’s founders—Xavier and Lucas Fischer—did not join Apple, but a number of their employees did.

Apple has reportedly been investing heavily to catch up to its rivals in the AI market, but is planning to keep its solution on-device, rather than relying on the cloud. Apple’s plans are not surprising, given the company’s long history of focusing on user privacy, but they do pose unique challenges deploying something as unique and computationally intensive as AI models.

Apple clearly sees Datakalab’s technology as an important piece of its future AI plans, and will no doubt leverage it to produce one of the most unique AI deployments yet.