Invest Like the Best• Mar 18, 2025• 1:02:15Interview

The AI Lab Revolutionizing Smell

From Invest Like the Best

Alex Wiltschko•Co-founder & CEO, Osmo

Executive Summary

Osmo is a company that has successfully digitized the sense of smell, creating an AI platform called "Olfactory Intelligence" (OI) to map, create, and reproduce scents from their molecular structure.
The company's technology was born from breakthroughs at Google Brain, particularly the application of graph neural networks to chemistry, which solved a century-old scientific problem of predicting a molecule's scent.
Osmo is commercializing its technology through a new fragrance house called "Generation," which uses AI to democratize and accelerate custom scent design, and by developing novel sensor applications like counterfeit detection for partners like StockX.
The long-term vision extends to significant healthcare and wellness applications, such as using scent analysis for non-invasive disease diagnostics.

12 quotes

Concerns Raised

Execution risk in scaling a novel, complex technology from the lab to mass-market products.
Building the necessary infrastructure, such as sensory panels, from scratch is capital and time-intensive.
Uncertainty around market adoption and the societal impact of giving computers a sense of smell.

Opportunities Identified

Disrupting the multi-billion dollar fragrance and flavor industry with a faster, AI-driven creation process.
Developing high-value B2B applications in product authentication, quality control, and materials science.
Pioneering a new field of digital health through non-invasive diagnostics based on analyzing volatile organic compounds (VOCs) in breath or sweat.

Key Themes

The Digitization of a Human Sense

Osmo has created a system that can "round trip" a smell: analyzing a physical scent, converting it into digital data, and then physically reconstituting it elsewhere. This breakthrough, demonstrated with their "Plum 1.0" scent, represents a fundamental step in giving computers a sense of smell, with profound implications for technology and society.

Digitizing a biological sense opens up entirely new categories of applications, from digital communication and entertainment to industrial quality control and diagnostics, that were previously impossible.

AI as a Tool for Scientific Discovery

The core of Osmo's technology is the application of advanced AI, specifically graph neural networks, to the complex domain of chemistry and olfaction. This approach enabled them to solve the long-standing problem of predicting scent from molecular structure, a feat validated in a double-blind trial where the AI outperformed human panelists.

This demonstrates how modern AI techniques are moving beyond data analysis to become powerful engines for fundamental scientific research, accelerating discovery in fields like materials science, drug discovery, and sensory neuroscience.

Full-Stack Innovation in a Legacy Industry

Instead of just selling software to incumbents, Osmo is building a vertically integrated fragrance house called Generation. By combining their proprietary AI with human perfumers, they aim to disrupt the traditional, slow, and exclusive fragrance industry by making custom scent design faster, more affordable, and accessible to a wider audience.

This full-stack strategy provides a powerful go-to-market approach for deep tech companies, allowing them to capture more value, control the user experience, and drive market adoption faster than by acting as a simple technology vendor.

Building a Proprietary Data Moat

Osmo recognized that off-the-shelf data was insufficient for their needs, so they built an entire internal ecosystem for generating olfactory data. This includes in-house sensory panels for human labeling, automated molecular sensors, and an active learning loop where the AI designs new molecules to be synthesized and smelled, continuously improving the models.

In specialized AI domains, the ability to generate unique, high-quality data is a critical competitive advantage. This case highlights that building the data generation engine can be as important as building the AI models themselves.

Get started free

Topics

Osmo AI Olfaction Sense of Smell Scent Digitization Graph Neural Networks Chemistry Neuroscience Fragrance Industry Perfumery Sensory Technology Counterfeit Detection Healthcare Diagnostics Business Model Innovation Google Brain

Processed Apr 3, 2026 yt-dlp + mlx-whisper + Gemini