Soul Verified
DataCareersBuy data
the founding table

three seats, with numbers on them.

soul verified is the marketplace for physical AI data, and our own industrial data comes from partner factories. three seats, every one with a real band and real equity below. we are pre revenue and honest about it: until the round closes, the base is lower and the equity is higher, and we will tell you exactly where we are on the first call rather than the third.

seat 01 · pune, india · full time

Capture Engineer

$80,000 to $120,0001.0% to 2.5% equity

own the rig and the floor. you instrument stations at partner factories with egocentric cameras and wrist sensors, read the line's own scan events where the line scans parts, and keep every stream synchronized across a shift.

what you own

  • the rig: camera mounts, sensor calibration, and the sync that aligns every stream
  • the episode slicer: segments raw sessions into episodes, using scan events as task boundaries where the line scans parts
  • relationships with factory floor managers -- the access IS the moat, and you maintain it

you have built hardware-software capture systems that ran in production, not just in a lab. you are comfortable on a factory floor and can hold a conversation with a line supervisor in Hindi or Marathi. Embedded systems, real-time sync, and sensor fusion.

apply for this seat →

seat 02 · new york or remote · full time

ML / Data Engineer

$140,000 to $180,0001.0% to 2.0% equity

own the pipeline from raw capture to a dataset a robotics lab can train on. every episode that ships in LeRobot v3.0 passed through your QA and your format validation.

what you own

  • the LeRobot v3.0 pipeline (with v2.1 import): Parquet and MP4, stream sync, episode labels
  • episode QA: reject corrupt frames, dropouts, out-of-sync streams, and anything a training run would choke on
  • Open X-Embodiment .zarr export, and whatever format the next lab asks for

you have worked with robotics data at scale. you know what a VLA training pipeline expects and what makes it fail silently. Python, PyTorch, Hugging Face datasets, and you have opinions about action representations.

apply for this seat →

seat 03 · new york · full time

Business Development

$110,000 base, $200,000 on target0.75% to 1.5% equity

sell datasets and evaluation runs to robotics labs. you know who is training what and which kinds of data their models need.

what you own

  • the pipeline from first conversation to a signed dataset licence
  • pricing -- per-episode, per-hour, and site-license models that make sense for both research labs and funded robotics companies
  • new capture environments -- every factory relationship you open is a new environment that makes the dataset more diverse and harder to replicate

you have sold technical products to research and engineering teams. you can hold a conversation about action chunking and VLA fine-tuning without a solutions engineer in the room. robotics, ML infrastructure, or data licensing background.

apply for this seat →
don't see your seat?

tell us what you have made.

if you would build this with us and your role isn't listed, reach out anyway. we care about what you have made, not your title.

krishi@achalalabs.com →
the real thing

build the dataset nobody else can.

three seats, real bands, real equity. the number this company is judged on is hours of verified production data shipped, and every one of these seats moves it.

krishi@achalalabs.com →
© 2026 Soul Verified. All rights reserved.
CareersPrivacyTermsDPAContact