Industries

Twelve sectors. One operating standard.

Every sector below is a live operating environment for us — sites we can access, participants we can recruit, and protocols we have run. Same standards, same delivery format, regardless of the industry.

Request a pilot See the tasks we capture
Manufacturing Logistics Agriculture Retail Healthcare Hospitality Construction Home Office Transportation Laboratory Education
01

Manufacturing

The most structured problem in physical AI — and the one where a policy's error budget is smallest.

Environments we access

  • Assembly and sub-assembly lines
  • Machine shops and press floors
  • Electronics and SMT lines
  • Textile and garment units
  • Pharma packaging halls

What makes it valuable

  • Two-handed assembly with strict sequence
  • Tool use under real torque and resistance
  • Defect calls made in under a second
  • Recovery from mis-feeds and jams

Typical sensor stack

Egocentric videoStereoWrist IMUAudio
02

Logistics

The highest-volume manipulation problem in the world, and the one with the least predictable object set.

Environments we access

  • E-commerce fulfilment centres
  • Cold-chain and ambient warehouses
  • Parcel sortation hubs
  • Quick-commerce dark stores
  • Truck docks and yards

What makes it valuable

  • Bin picking from genuinely mixed SKUs
  • Carton handling at commercial speed
  • Deformable packaging and polybags
  • Long multi-item picking sequences

Typical sensor stack

Egocentric videoDepthIMUGPS (yard)
03

Agriculture

Deformable, variable, living objects under lighting no indoor dataset contains.

Environments we access

  • Open fields and smallholder plots
  • Greenhouses and polyhouses
  • Orchards and plantations
  • Packhouses and grading lines
  • Dairy sheds

What makes it valuable

  • Delicate grasping without bruising
  • Ripeness and quality judgement
  • Full daylight range, dust and monsoon
  • Occlusion-heavy scenes in dense foliage

Typical sensor stack

Egocentric videoStereoGPSAudio
04

Retail

Dense shelving, thousands of package geometries, and members of the public in the workspace.

Environments we access

  • Supermarkets and hypermarkets
  • Apparel and footwear stores
  • Pharmacies and convenience formats
  • Dark stores and stockrooms
  • Traditional kirana and wet markets

What makes it valuable

  • Shelf replenishment in tight clearances
  • Fragile and irregular packaging
  • Operating with customers present
  • Constant SKU turnover

Typical sensor stack

Egocentric videoDepthAudio
05

Healthcare

Protocol-bound work where sequence matters as much as motion, and error is not recoverable.

Environments we access

  • Hospital wards and supply rooms
  • Diagnostic and pathology labs
  • Sterile processing departments
  • Hospital and retail pharmacies

What makes it valuable

  • Strictly ordered, auditable sequences
  • Sterile-field handling constraints
  • Counting and verification behaviour
  • Rare-event and exception handling

Typical sensor stack

Egocentric videoStereoAudioConsented & de-identified
06

Hospitality

Heat, liquids, deformables and time pressure — the hardest combination in service robotics.

Environments we access

  • Hotel and banquet kitchens
  • Cloud kitchens and QSR chains
  • Cafés and beverage counters
  • Housekeeping floors

What makes it valuable

  • Cooking judgement under time pressure
  • Plating consistency across repetitions
  • Liquid pouring and transfer
  • Room turnaround as a long-horizon task

Typical sensor stack

Egocentric videoStereoIMUAudio
07

Construction

A worksite that is different every single day, with heavy payloads and no fixtures.

Environments we access

  • Residential and commercial builds
  • Infrastructure and road projects
  • Prefabrication yards
  • Interior fit-out sites

What makes it valuable

  • Heavy, awkward material handling
  • Tool work on uneven, unstable ground
  • Layout, marking and measurement
  • Scene that changes hour to hour

Typical sensor stack

Egocentric videoDepthIMUGPS
08

Home

The benchmark environment for humanoids, and the one that cannot be simulated convincingly.

Environments we access

  • Urban apartments across layouts
  • Independent houses
  • Rural and semi-urban households
  • Shared and multi-generational homes

What makes it valuable

  • No two homes are laid out alike
  • Long-horizon, multi-room tasks
  • Cloth, food and clutter together
  • Genuine variation in tools and utensils

Typical sensor stack

Egocentric videoStereoDepth
09

Office

Shared human space, where a robot's job is as much about not getting in the way as about the task.

Environments we access

  • Corporate campuses
  • Coworking and shared offices
  • Institutional and government buildings
  • Mailrooms and facility stores

What makes it valuable

  • Continuous operation around people
  • Paper and thin-object manipulation
  • Lifts, doors and access control
  • Predictable layout, unpredictable traffic

Typical sensor stack

Egocentric videoDepthAudio
10

Transportation

Confined access, high torque, and the last fifty metres that no autonomous vehicle solves.

Environments we access

  • Service workshops and tyre bays
  • Fleet and bus depots
  • Ports and container yards
  • Last-mile delivery routes

What makes it valuable

  • Working inside tight, awkward spaces
  • High-torque tool use
  • Doorstep handover to a person
  • Stairs, gates, lifts and gatekeepers

Typical sensor stack

Egocentric videoStereoIMUGPS
11

Laboratory

Millimetre precision on small, fragile objects, executed the same way every single time.

Environments we access

  • Commercial testing laboratories
  • Corporate R&D centres
  • Quality-control facilities
  • University and institute labs

What makes it valuable

  • Fine bimanual precision at small scale
  • Protocols that never vary
  • Transparent and reflective glassware
  • Expert gaze as a supervisory signal

Typical sensor stack

Egocentric videoStereoEye tracking
12

Education

Where skilled trades are taught — and where the same task exists in expert and novice form side by side.

Environments we access

  • Industrial training institutes
  • Polytechnics and vocational centres
  • University engineering workshops
  • Skill-development programmes

What makes it valuable

  • Instructor demonstrations, deliberately clear
  • Expert and novice attempts at one task
  • High repetition of the same procedure
  • Failure modes captured, not edited out

Typical sensor stack

Egocentric videoStereoAudio narration

Working in a sector that isn't listed?

These are the sectors where we already have site access and a participant network. If yours is not here, tell us what you need and we will scope the access for it.

Schedule a Call Browse use cases