The foundational AI model for food preparation
ChefOS is a generalized food manipulation model trained on tens of millions of real-world servings.
Fleet Learning
Chef's systems continually improve the more picks they do.
Flexible
Our robots work with almost any ingredient, no matter size, stickiness, deformability, or wetness.
Pre-Trained
Chef's systems have been trained on tens of thousands of hours of production.
Any Portion Size
Chef can work with any portion size of ingredient.
Any Tray, Cup, or Burrito
We work with your containers, and can target any compartment or placement within them.
Embodied AI
We believe AI’s most powerful impact will be on the physical world. The physical world represents 90% of GDP. ChefOS is built on modern advancements in AI like foundational models, learning from demonstration (imitation learning), and deep learning.
ChefOS is to the physical world what ChatGPT is to language. GPT predicts the next word in a sequence. ChefOS’s Generalized Food Manipulation Model knows which robot action should come next in the real world.
01. See
02. Think
03. Act
Your browser does not support the video tag.
.webp)
Chef has been trained on tens of thousands of hours of production data
116,853,746
Servings in Production
The Data Engine Flywheel Continually Improves Customer Value
.png)
Chef has consistently picked thousands of ingredients in production.
Chef has placed on to dozens of conveyors and placement quality improves with every pick.
Embodied AI’s performance is based on quality and raw quantity data from the physical world. The AI System that is trained on the most and best data will simply perform the best.
This is especially true for food since food is so highly variable; the cheese grits a kitchen makes one day are a bit different the next day. The cheese grits at the West Coast facility are a bit different than at the East Coast facility. And the cheese grits across different food producers are different entirely.
Every ingredient is so radically different that vast amounts of data from robots in production are needed to train our food manipulation model to be able to reliably, flexibly, and consistently manipulate any ingredient no matter how it’s cooked or cut all while not damaging it, meeting throughput, placing into the right part of a tray, and not spilling in transit.
Improvement Over Time
By leveraging modern advancements in AI, ChefOS can deal with the variance…
Between different ingredients
Across different batches of the same ingredient
What does this mean for you?
High consistency and high yield, even in high-mix kitchens.
One AI Platform for All Your Needs
Trained on millions of deposits from ChefOS enabled systems, the ChefOS enables your robots to pick almost any ingredient from first day of deployment.
Interact with ChefOS
Ingredient in 3D
Mixed Veggies
Pulled Meat
Leafy Greens
Single Pieces
Fresh, cooked, and IQF “scoopable” food
Sauces, mashes, chilis, curries, stews
Piece picking from unstructured totes
How It Works
ChefOS dynamically picks based on the particular ingredient and its properties, the pan topography, how the ingredient was cooked and prepped on that particular day.
ChefOS detects and tracks trays as they move down the line. Chef can re-train these models to place into any tray, into any compartment, and on any conveyor.
ChefOS's Applications
ChefOS powers all of Chef’s applications in every industry
Learn How We've Helped Our Customers
CASE STUDIES
Project Open Hand
The San Francisco-based nonprofit uses Chef robots powered by physical AI to address staff shortages and deliver consistent, medically tailored meals to people in need.
Cafe Spice
Using Chef’s robots, Cafe Spice was able to boost output by 2-3x, ease labor shortage constraints, and lower food giveaway by 67%.
Amy’s Kitchen
Beyond improving consistency, the company increased labor productivity by 17% and reduced food giveaway by 4%.