NOUS
Nous
For AI labs

Training data and RL environments for AI labs.

For agents that act: on the phone, in the browser, and in the physical world.

Request a sample Talk to us
  1. 01Voice & CX agent environments
  2. 02Website & app replica environments
  3. 03Egocentric multi-sensor video
  4. 04Action-labelled gameplay
  5. 05Robotics teleop data
What we deliver

Five tracks. Ask for a sample of any of them.

  1. 01
    Voice & CX agent environments RL environments Live stack

    Graded tasks for voice agents, run over real phone calls on Kelam, our production voice platform.

    You get
    • Tasks with rubrics and graders
    • IVRs, interruptions, EN/ES/FR mid-call, tool use
    • Audio, transcript and tool trace per episode
    Checked by
    Graders compared against hand-scored calls.

    Calls are recorded. Textable too — SMS Terms.

    Request a sample
  2. 02
    Website & app replica environments RL environments

    Resettable replicas of real web apps for computer-use and browser agents.

    You get
    • A containerised replica with seeded data
    • Tasks graded on the app’s final state
    • Optional human demonstrations
    Checked by
    Graders tested against wrong and partial solutions.
    Request a sample
  3. 03
    Egocentric multi-sensor video Dataset

    First-person recordings of hands-on tasks, captured in India with consented, paid participants.

    You get
    • Head and wrist cam, time-synced
    • Depth and tactile where the task needs it
    • Task and action-segment labels
    Checked by
    Training a policy on it, not eyeballing it.
    Request a sample
  4. 04
    Action-labelled gameplay Dataset

    Screen recordings with synced keyboard and mouse input, from partner gaming cafés in India.

    You get
    • Frame-aligned input logs
    • Game, version and settings per session
    Checked by
    Alignment measured on every session; drifted sessions are cut.
    Request a sample
  5. 05
    Robotics teleop data Dataset Scoped per lab

    Teleoperated manipulation demos. Tell us the embodiment and tasks; we agree the setup before recording.

    Talk to us
Built by us

Founded by Kaushik Kanna and Sami Sadfa, both previously at Giga (YC '23) and Microsoft. The engineers who scope your environment are the ones who build it.