Based in the San Francisco Bay Area, our founders have built and deployed real-world AI trusted by U.S. federal agencies. We bring that same engineering rigor to custom, sensor-rich training data for physical AI.
Human demonstrations.
Your sensors. Your specifications.
Real tasks are messy. Movement is nuanced. Contact changes everything. Models need more than video—they need the physical context behind every action.
We capture that context at the source: people performing purposeful tasks in custom-designed, sensor-equipped suits.
You define the tasks, modalities, and sensor requirements. We design the wearable setup and collection protocol to capture the signals that matter.
Explore capture modalitiesSilicon Valley veterans with hands-on experience bringing enterprise and physical AI into Fortune 100 companies, U.S. government departments, military applications, and international governments, including the UK.
Meet the experience behind the dataA capture stack shaped by your research.
Choose the modalities. Define the detail.
Capture the human demonstrations your policy needs: full-body movement, fine motor actions, and task transitions, using a suit and sensor layout designed around your research.
Every engagement is custom. Sensor configurations, capture targets, and quality criteria are agreed with your team.
Research-led from the first conversation
to the final handoff.
We work with your researchers to understand the target behaviors, environments, sensor needs, and acceptance criteria.
We design the human-worn suit, integrate the agreed sensors, and develop task-specific demonstration protocols.
Human demonstrations are collected and reviewed against your requirements. Pilot feedback shapes the collection before scaling.
Receive datasets in agreed formats, with the documentation and context your team needs to put them to work.
Our founders’ experience spans the entire stack—from inventing sensors to deploying intelligence in environments where performance matters.
Built real-time camera and sensor acquisition interfaces at MATLAB, connecting physical signals to the tools engineers use to understand them.
Invented low-cost multispectral sensors to help prevent fraud in financial transactions—bringing hardware innovation and applied intelligence together.
Developed underwater physical AI systems to precisely estimate fish morphology, working at the intersection of sensing, perception, and difficult real-world conditions.
Co-founded and developed a video redaction and face search AI SDK used to support investigation workflows at large law-enforcement and federal agencies.
Deployed low-latency AI for metal recycling and sorting, combining multimodal sensor inputs to make decisions under the constraints of a live industrial environment.
Tell us what your model needs to learn.
We’ll design the data to get it there.