Aryballe

Where Osmo predicts smell from a molecule's structure, Aryballe goes at the problem from the other end: it builds a device that sniffs. The company is based in Grenoble, France's optics and microelectronics cluster, and its instruments borrow from biology - peptides that bind odour molecules, read optically, with the resulting signature interpreted by machine learning. The output is a repeatable fingerprint of a smell rather than a human's description of it.

The reason this belongs in a fragrance directory is quality control. Perfumery has always leaned on trained human panels, and panels are expensive, slow, and inconsistent by the afternoon. A digital nose does not replace a perfumer's judgement, but it can tell you whether batch 400 smells like batch 1, whether a plastic has gone off, or whether a car interior has the odour profile the brand signed off. Aryballe's customers sit across automotive, food and drink, cosmetics and healthcare - which is a good indication that this is production kit, not a demo.

It is also a useful corrective to the idea that scent tech is all generative AI. Aryballe has been shipping sensor hardware for years and has taken investment from strategic industrial backers including Samsung Venture Investment and Asahi Kasei - the kind of money that shows up when a thing works on a factory floor.

This is business to business and there is nothing here for a consumer to buy. Read it as one half of the digital-olfaction picture: Osmo models what a molecule will smell like, Aryballe measures what is actually in the room.