TensorTest GTM brain

TensorTest builds creative-work evaluations, reinforcement-learning environments and expert training data for AI labs developing agents.

category

Creative agent evaluation and training data

market profile

  • text: Creative agents can claim success despite failing the actual deliverable, making outcome-based evaluation valuable.
  • text: Expert data vendors increasingly offer evaluations, environments and training data together.

personas

  • title: AI lab evaluation lead
    pains:
    • Self-reported agent success does not establish whether a creative brief was fulfilled
    • Human creative judgment is expensive and hard to turn into repeatable evaluation
  • title: Post-training and reinforcement learning research lead
    pains:
    • Needs task environments and reward signals that reflect professional craft
    • Needs demonstrations and preference data for weaknesses found in evals

icp

  • text: Foundation model or agent labs building creative capabilities and evaluating unreleased checkpoints.
  • text: Research teams with video editing or 3D workflows and a budget for commissioned evaluation and expert data.

message pillars

  • text: Evaluate finished professional work, not an agent's claims.
  • text: Use complete creative jobs as the unit of measurement.
  • text: Anchor quality tests in professional judgment.

objections

  • objection: Can an automated verifier judge subjective creative quality?
    response: Examine calibration against blind professional judgments and test failure cases on a private evaluation.
  • objection: Will our unreleased checkpoint and results remain private?
    response: The site offers private evaluations; confirm access, retention and contractual terms before a run.

competitors

  • name: Mercor
    website: https://www.mercor.com/research/
    why: Provides expert-built benchmarks, evaluation environments and datasets.
  • name: Scale AI
    website: https://scale.com/off-the-shelf-data
    why: Offers expert training data, evaluations and reinforcement learning environments.

keywords

  • creative agent benchmark
  • video editing AI evaluation
  • agent reinforcement learning environments
  • expert creative training data
  • video editing agent benchmark
  • AI agent verifier

voice

  • research-led
  • specific and evidence-oriented
  • plainspoken

signals

  • text: A lab is preparing a creative agent or multimodal checkpoint for release.
  • text: A research team is hiring for agent evaluations, reward modelling or creative data.

stack

  • Blender

pricing

model: unknown

motion

sales_led

channels

  • research publications
  • benchmarks
  • Hugging Face
  • direct sales
  • X
  • LinkedIn

confidence

medium

Founders

  • name: Pavan Kalyan
    title: Founder/CEO
    role: CEO
    bio: Cofounder and CEO @ TensorTest. Prev. @ Microsoft Research focused on training language models. Graduated from IIT Bombay with Btech + Mtech in Electrical+AI. Published @ Neurips, ACL, Interspeech.
  • name: Nirmit Arora
    title: Founder/CTO
    role: CTO
    bio: Co-founder and CTO @ Ritivel. Prev @ Microsoft Research focused on making agentic systems safe.
  • name: Gunin Gupta
    title: Founder/COO
    role: COO
    bio: Cofounder and CBO @ Ritivel. Prev. @ AT Kearney focused on sustainability and energy process industries. Graduated with a bachelor's from IIT Bombay.

Back to YC companies