Artly's software platform allows robots to perform complex tasks, with or without demonstrations, using various hardware.
Artly General Intelligence (AGI) Platform
Artly provides a generalized training and learning system for humanoids, a critical software layer to train, refine, and scale humanoid intelligence, making robots of tomorrow master practical skills to be helpful from day one.
This AI-first approach solves the critical data issue, making humanoids scalable for commercial jobs.
Growing list of skills learnt from human experts
Artly offers many basic skills out of the box and allow user to fine-tune or create new skills. We’re building a platform for intelligent, general-purpose manipulation. Our goal is to create robots that can grasp, pick, cut, pour, stir, and serve—all with the grace and precision of a human hand. It started with the Barista Bot. We trained it to grind beans, tamp espresso, steam milk, and pour drinks—all tasks that require precise, repeatable movements. But as we advanced our hardware and AI platform, we began expanding beyond coffee. Now, Artly’s robots are learning new physical skills like:
* Grasping a wide range of objects – From small sugar packets to slippery ceramic cups, our robots can adapt their grip using tactile sensors and vision-based control.
* Picking and placing – Whether it’s a spoon, milk pitcher, or a garnish, our system can identify, pick up, and precisely position items in context.
* Cutting and slicing – With safety and control at the core, we’re training robots to handle knives and blades to slice fruit, cut packaging, and prepare ingredients—tasks that once seemed out of reach for robotic arms.
* Stirring, pouring, and serving – Our robots can now stir drinks, melt chocolate, and garnish cups, creating more complex, interactive beverage experiences.
From Intent to Action: Task Composition with Artly.
To achieve this, Artly uses a large language model (LLM) to automatically generate the conditions and sequence of actions required to fulfill a user’s goal. When a user provides a high-level instruction—like “make a fruit salad”—the system breaks it down into structured steps by referencing pre-stored skill sets in the Artly platform. The LLM composes a logical plan, including preconditions, task order, and tool usage, allowing the robot to execute complex, multi-step tasks with minimal human input. This approach enables flexible, goal-driven behavior across a wide range of real-world scenarios.
Recipe description: “Please prepare me a fruit salad with the fruit on the table.”
Locate the fruit on the table. Locate cutting board and knife. Locate Bowl.
Pick individual grapes and place it into the bowl.
Slice the oranges and then put the slices inside the bowl.
Cut the apple in half and put two halves inside the bowl.
“Your salad of grapes, oranges, and apples are inside the bowl.”
Pre-condition: there is a fruit knife within right arm’s reach. There is a wood plate and bow within left arm’s reach
Cut the apple in half: Use your left hand to find the apple inside the wood plate. Pick one of the apples. Move the apple to the cutting board. Use your right hand to grab a fruit knife and cut the apple. Then put the knife back to its original place. Put the two halves to the bow
Post-condition: the two halves of apple are inside the bow.
