Animal motion for generative video
Video models need to get the motion right, from a horse changing gait to a bird folding its wings. We build annotated examples, expert reviews and benchmarks that make movement errors easier to find and fix.
Data infrastructure
for the living world.
We’re building multimodal training data for AI
to understand and operate in biological environments.
Living systems
never stand still.
Growth, behaviour and changing conditions shape every interaction. Understanding them takes more than a snapshot.
Capture what an expert sees.
What they see, what they infer, and what they do. In one connected record.
Training data and benchmarks
for the living world.
We collect observations, label what matters, and record expert judgments and the reasoning behind them. We turn that work into training data, evaluations and benchmarks that help teams improve their AI models.
Gecko-inspired adhesion and materials
New adhesives and silk-inspired fibres start with understanding what gives the natural material its properties. We combine measurements, expert explanations and test results into datasets and evaluations for teams building materials models.
Expert POV in field environments
Field skills are often missing from a written protocol. We record experts doing the work, annotate the steps and capture the decisions they make along the way. Those records become training demonstrations and tests for physical-AI systems.
Peregrine flight for aerial robotics
We help aerial robotics teams turn animal flight into something their models can learn from. Synchronized video, motion measurements and expert analysis feed datasets and benchmarks for studying new approaches to drone navigation and control.
Laboratory animal behaviour
To study how an animal responds to a new drug, researchers need behaviour, context and a consistent way to assess change. We bring those records and expert judgments together in datasets and benchmarks for preclinical AI.
Robotic harvesting without bruising
Picking robots need to learn which movements and forces damage a crop. We connect each attempt to the fruit’s condition afterwards, add expert labels and explanations, and build evaluations that test both picking success and damage.
Livestock health and welfare
For livestock-monitoring systems, we build data around the decisions farmers and vets need to make. Movement recordings, independent assessments and expert reasoning become labelled examples and benchmarks for detecting changes that need attention.
Pet behaviour monitoring
For humanoid robots to share a home with pets, they need to anticipate how an animal may respond. We combine video, behaviourist assessments and observed outcomes into datasets and benchmarks for recognising cues and testing predictions.
Fish feeding and behaviour
We help aquaculture teams connect what happens in the water to feeding and monitoring decisions. Video, feed logs and water measurements are annotated with expert judgments, then used to build training examples and evaluations for their models.
The next generation of AI
will need to understand
life itself.
Tell us what your model needs to learn.
Help shape the data it learns from.