Osmo is pushing AI into one of the most complex human senses: smell. In a New Jersey lab, trained human sniffers are cataloging odors with descriptors like “nutty,” “peachy,” and “marine,” while the company pairs those observations with chemical data to train what it describes as a foundation model for scent.
The startup says its AI “nose” is now as good as humans at identifying and describing smells. That is a meaningful milestone because odor is notoriously difficult to digitize: tiny molecular differences can produce dramatically different scent experiences, and human perception is subjective.
Why this matters
If systems like Osmo’s continue to improve, they could make smell measurable, searchable, and programmable in ways that were not previously possible. That could benefit fragrance and flavor development, quality control, environmental monitoring, and potentially even medical diagnostics where disease can change a person’s odor signature.
- Better products: AI could help create more appealing perfumes, foods, and household goods.
- Faster discovery: Models may reduce the trial-and-error process of scent design.
- New sensing tools: Digital odor detection could support health, safety, and industrial applications.
While the technology still needs broader validation and real-world deployment, Osmo’s progress points to an exciting future where AI can help decode an entire sensory world that computers have barely been able to understand.