Waddle Labs
Waddle Labs is an applied AI lab building LLM-based agents that control robots in real time.
The Problem
Robots today struggle to operate outside carefully prepared settings, and making a robot useful typically requires work specific to one machine, task, or lab. This makes progress hard to reproduce or extend, and risks concentrating robot intelligence in the hands of a few large companies. Individuals and smaller labs lack ways to adapt robots to their own needs without this bespoke, hard-to-transfer engineering effort.
The Solution
Waddle Labs builds LLM agents that control robots and learn in context, with the company observing its agents become more reliable over time through use. Its research focuses on three areas: preserving and transferring what one agent learns to another and building shared memory across agents, drawing on continual learning research; improving inference efficiency so agents become faster with practice while remaining flexible to new tasks; and designing human-robot interaction so people can assign tasks, observe robot behavior, and correct mistakes safely. The company states it has shown language models solving robotics tasks in real time on real robots.