Firsthand is piloting SkillCap in Chinatown kitchens
Request footage
SkillCap, a black cap with a camera housing built into the brim, above a glowing planet horizon

The camera that teaches robots real work

  • Hands only
  • Faces blurred on site
  • Paid per hour
  • Consent first

Chinatown pilot

For robot labs

Request footage

Skill
you own.

SkillCap is a camera built into a work cap. Cooks, folders and mechanics wear it on a normal shift. Their hands become training data for robots, and they are paid for every hour that passes review.

View SkillCap

Inside SkillCap

SkillCap assembled SkillCap taken apart: cap, brim, housing shells, the KC-1 board, motion sensor, camera module, cable and power bank
Where SkillCap works
SitesChinese kitchensBanh mi shopsLaundromatsAuto shops

SkillCap parts

  1. The custom wide-angle camera module on its flex cable
    Camera
    Our wide-angle camera module
    Points down at the hands
  2. KC-1, the custom camera board inside the brim housing
    Compute
    KC-1, our own board
    Encodes on the chip, streams to the shop box
  3. The motion sensor on the KC-1 board
    Motion
    Head IMU, on the board
    On the same clock as the video
  4. The shop box, a small aluminium computer that stays in the back room
    Shop box
    Edge computer with an AI chip
    Blurs faces before saving
A cut-away illustration of a small restaurant kitchen with the shop box on a shelf

Faces stay in the shop.

The cap streams to a box in the back room. It blurs faces and screens, drops the audio, then saves. Nothing leaves the shop until it is clean.

A shift, cut into skills.

Each session is split into tasks in the order they happened, so a robot learns the whole flow of real work, not one staged move.

  1. Wash and trim bok choyChop scallion, finePortion pork for char siu
  2. Wok toss, high heatPlate and wipe downLadle soup into bowls
  3. Scrub the wokWrap and label leftoversWipe the steel counters

No audio

Delete on request

The view from SkillCap: the wearer's hands chopping scallions with a cleaver

This is what a robot learns from.

Hands, tools and food, from the worker’s own view, every step labelled. Sample labels shown.

View from SkillCap

Ready for robot labs.

LeRobot WebDataset
SessionSiteTaskClean
example-01KitchenChop scallionYes
example-02KitchenWok tossYes
example-03Banh miSplit baguetteYes
example-04LaundryFold towelReview
Sample sessions
The shop box
Chop scallion, 3 segments

Raw

1080p video and head motion on one clock, with a JSON record per session.

Cleaned

Faces blurred, lens calibrated, cut by task and scored for quality.

Labeled

Task steps, hand keypoints and object masks, as a LeRobot dataset.

Where SkillCap works

SkillCap
Wok
Cleaver
Steamer
Chopsticks

Learns knife, wok and steamer work from Chinatown kitchens.

Baguette
Bread knife
Board
Sauce

Learns fast bread and filling work from banh mi shops.

Towels
Washer
Hanger
Basket

Learns folding at real speed from wash-and-fold laundromats.

A humanoid robot's hands chopping scallions in a small restaurant kitchen

Every robot needs a shifu.

The robots coming into kitchens and laundromats will learn from people. Firsthand makes sure those people are asked first, paid fairly and credited.

Concept render

Own your skill.

Request footage