BRENNEN

Industries / Food, farming and forestry

Flavors and fragrances

The flavors and fragrances industry designs and makes the tastes added to food and drink and the scents added to perfume, cosmetics, soap and cleaning products, and sells them to the brands that make those goods.

1768

Antoine Chiris founded a house in Grasse, in the south of France, that turned flowers into raw materials for perfume; France's heritage register gives the founding year, and the flavor and fragrance house Givaudan dates its own roots to it.

pop.culture.gouv.frgivaudan.com

38

Milestones achieved across 6 domains.

11

To watch for: announced, not yet confirmed outside.

12

Open deliverables. Nobody has shipped them.

In one minute

Where it started

Perfumers around Grasse, a leather-tanning town in the south of France, have turned flowers into scented materials since at least the 16th century, and in 1768 Antoine Chiris made a business of supplying such materials to others. Chemists then began to make scent molecules in the laboratory: coumarin in 1868, vanillin in 1874 and a violet scent in 1893. In 1882 Houbigant's Fougère Royale was built around a lab-made molecule, and perfumes and food flavors have mixed natural and lab-made materials ever since.

Where it stands

A few large houses, among them Givaudan, IFF, dsm-firmenich and Symrise, compete for "briefs", the requests in which a brand asks several suppliers for a flavor or scent. Competition authorities in Europe and Britain have investigated suspected collusion among the largest houses since 2023, and in 2024 the European Commission fined IFF €15.9 million after an employee deleted messages exchanged with a competitor. Natural materials still follow harvests, storms and fires, and the rules keep narrowing what may be used: the EU banned the floral scent Lilial from cosmetics in 2022 and removed every smoke flavoring from its approved list in 2024.

What is open

Most of what the trade knows sits in private archives: formulas, panel ratings and the list of which plant makes which molecule. Open jobs below include a robot that pollinates vanilla flowers, an open set of perfume formulas taken from expired patents, a model that predicts when the two mirror-image forms of a molecule smell different, and an independent test of the AI "noses" that start-ups say now match people. Others are test sets and registers: formulas with checked answers for the calculators that apply IFRA's limits, and a register of the plants that make each aroma chemical, which would show where a single fire can stop supply.

Domains

AchievedWatch forOpen

Open in flavors and fragrances

3 of 12 open deliverables

Open · Natural ingredients

Robot pollinates vanilla flowers

A robot arm with a camera that finds open vanilla flowers on a vine, lifts the small flap that keeps the flower from pollinating itself, and presses the pollen onto the stigma, as hand pollinators do with a toothpick.

Built with what exists

A low-cost open robot arm trained with Hugging Face's LeRobot software (see the Note "No download"); a depth camera; an RF-DETR object detector trained on the open Roboflow flower set; a 3D-printed tip shaped like the pollinator's toothpick.

Credit · unclaimed

Open · Making aroma chemicals

Mirror-image smells predicted

An open test set of at least 150 pairs of enantiomers, the two mirror-image forms of one odor molecule, each with published odor descriptions and, where known, odor thresholds, plus a model that predicts for each pair whether the two forms smell alike or different, and how.

Built with what exists

the public Leffingwell "Chirality & Odour Perception" web pages and the papers they cite; RDKit to build 3D shapes of each molecule; Uni-Mol (MIT licence) and OpenPOM (MIT licence) as starting models.

Credit · unclaimed

Open · Creating a flavor or scent

Perfume formulas from old patents

An open set of complete fragrance formulas, each ingredient with its amount, taken from the worked examples in expired fragrance patents and tagged with the scent family each patent gives, such as citrus, floral or woody, plus a model that predicts the family from the ingredient list alone.

Built with what exists

Full-text patents from Google Patents and the USPTO's bulk data, searched in the patent classes for perfumes (C11B 9/00) and perfume preparations (A61Q 13/00); a language model such as Claude to pull the formula tables out of patent text, with a person checking each one (the Note "Seven more flagged" covers a Claude-based reviewer of documents); PubChem's free API to match ingredient names to structures; RDKit; a gradient-boosting classifier as a first model.

Credit · unclaimed