Data, made to order.

VectorPax records how skilled people do real work, so machines can learn it from them with consent, with credit, and with pay.

Halftone dot illustration of a chef's hands dicing herbs and an onion on a plate

The next gains come from expert physical skill. It cannot be scraped from the internet.

Halftone dot illustration of a chef's hand reaching down, with a lime tracking dot on each fingertip

Robots must learn physical tasks, and that data is not on the web. Ordinary first-person video is now a glut. Labs need expert skill, captured well.

Work with us

Your robots learn from video. Most of it shows amateurs.

Halftone dot illustration of an onion on a cutting board, drawn as concentric depth contour rings
  • Professional technique.

    Knife work, sauces, plating, and timing across stations. Trained chefs, not home cooks.

  • Richer capture.

    First-person video plus depth and hand motion, with every action labeled.

Available now

  • Kitchen Signal

    First-party engagement signals from a food and lifestyle audience, across recipes, cooking, food history, and diet content.

  • The Culinary Dataset

    Professional recipe video, rights cleared, ready to license.

All datasets
Halftone dot illustration of a blank order ticket feeding out of a printer slot

A lab sends a spec. We capture, label, and deliver.


  1. Tell us the task.

    Tell us what kitchen task you want captured next, and in what format.

  2. We capture.

    Professional chefs record the work to your written spec.

  3. We label and check.

    One finished hour is captured, labeled, checked, and ready to deliver.

  4. We deliver.

    Every dataset carries documented consent and ownership.

Exclusive

A higher price.

The buyer holds sole rights for a set term.

Shared

A lower price.

We keep the right to license the data again.

Cooking is hard for robots.

Heat, knives, soft ingredients, and timing make it one of the hardest jobs to learn.


We sell what a crowd cannot make. We do not compete with phone footage of people cooking dinner.

Halftone dot illustration of a pan searing food over a ring of burner flames

A guild of skilled hands. Chefs first, athletes next.

Halftone dot illustration of six chef's hands at work: seasoning, holding a knife, folding cloth, whisking, plating with tweezers and holding an egg
  • Members record expert work to a buyer’s written spec. Labs license the data. Members own it and share in every license, including future ones.

  • The buyer always pays. Chefs never pay us.

  • AI learns from experts who are paid and credited, not from data taken without asking.

Rights first. License, never sell.

  1. Rights first.

    Every dataset carries documented consent and ownership before any buyer sees a sample.

  2. License, never sell.

    Every license names the buyer, the use, and the term.

  3. Never publish raw data.

    Data captured privately cannot be scraped.

  4. Owners get paid on every deal.

    We publish the split. Members see every deal they are part of.

  5. No made-up badges.

    We claim only standards and guarantees we can back up.

Data suppliers. Kitchen robot companies. Robot foundation model labs. Humanoid robot makers.

They need specialty capture they cannot do themselves.

Their robots must learn new dishes and tasks.

Work with us.

Tell us what kitchen task you want captured next, and in what format.


Opens your mail app