Markov GTM brain

Markov supplies expert computer-use datasets and CAD reinforcement learning environments to AI teams training and evaluating agents.

category

AI training data and RL environments

market profile

  • text: Computer-use agents need action-aligned demonstrations and evaluable tasks across real applications.
  • text: Markov combines public datasets with commercial off-the-shelf and custom data collection.

personas

  • title: Head of agent research
    pains:
    • Generic screen recordings lack synchronized actions and expert reasoning.
    • Training agents on professional CAD work requires scarce domain expertise.
  • title: ML data acquisition lead
    pains:
    • Collecting and labeling realistic workflows at scale is slow.
    • Evaluations need task briefs, gold outputs and consistent rubrics.

icp

  • text: Frontier model labs and well-funded AI teams building desktop or computer-use agents are the clearest target accounts.
  • text: Teams targeting engineering and design applications with CAD expertise needs are a strong fit; account size is not specified.

message pillars

  • text: Capture expert work as training data and environments.
  • text: Deliver workflows with action labels and narration.
  • text: Support buying ready-made data or commissioning a collection.

objections

  • objection: Will the data reflect our target software and tasks?
    response: Review a representative sample and discuss custom collection for the required workflows.
  • objection: Can we evaluate outcomes, not just imitate clicks?
    response: CAD environments include briefs, input files, gold outputs and rubrics; validate coverage against your own evaluation set.

competitors

  • name: AfterQuery
    website: https://afterquery.com
    why: Provides expert-sourced training data for AI labs.
  • name: Cua
    website: https://cua.ai
    why: Competes for computer-use agent training and evaluation workflows.
  • name: Datoric
    website: https://datoric.com
    why: Identified as a computer-use dataset competitor in market research.

keywords

  • computer use training data
  • CAD agent datasets
  • computer use action labels
  • reinforcement learning environments CAD
  • desktop agent benchmark
  • expert demonstration datasets

voice

  • Technical and specific
  • Evidence-led
  • Direct

signals

  • text: A lab is recruiting computer-use agent researchers or releasing desktop-agent benchmarks.
  • text: A team is building CAD automation and needs expert demonstrations or task-based evaluation.

stack

  • AutoCAD
  • Hugging Face

pricing

model: unknown

motion

hybrid

channels

  • Hugging Face
  • Research publications
  • X
  • LinkedIn
  • Direct sales

confidence

high

Founders

  • name: Dev Mandal
    title: Founder/CEO
    role: CEO
    bio: Co-Founder & CEO at Markov. Prev at Sarvam, Aerospace engineering at IIT Madras
  • name: Harish Ashok
    title: Founder/CTO
    role: CTO
    bio: Co-Founder & CTO at Markov. Previously founded Zenith, a robotics company building AI-assisted hardware tools.

Back to YC companies