Shofo GTM brain

Shofo supplies multimodal training datasets to AI labs building models that understand video, audio and computer use.

category

Multimodal AI training data

market profile

  • text: Frontier labs need specialized data beyond generic web text for multimodal and computer-use models.
  • text: Dataset sourcing competes with open datasets and managed custom data providers; licensing and coverage matter alongside scale.

personas

  • title: Head of multimodal data
    pains:
    • Sourcing enough footage with the right subjects and usable rights
    • Comparing coverage and transcript quality before committing training compute
  • title: Computer-use model research lead
    pains:
    • Finding expert demonstrations with aligned screen, keyboard and mouse events
    • Gathering enough varied workflows to improve agent generalization

icp

  • text: AI foundation model labs training multimodal or computer-use models, typically with dedicated research and data infrastructure teams.
  • text: Teams with substantial data requirements across video, speech, gameplay or expert work, rather than one-off annotation projects.

message pillars

  • text: Large multimodal supply for model development.
  • text: Expert computer-use demonstrations include aligned interaction logs.
  • text: Speech footage includes transcripts and speaker variety.

objections

  • objection: Does the data cover our exact task and have appropriate rights?
    response: Request a sample and review provenance, permissions and coverage before contracting.
  • objection: Can an open dataset meet the need at lower cost?
    response: Compare a sample on task fit, modalities, input logs and licensing against the open alternative.

competitors

  • name: TornadoLabs
    website: https://huggingface.co/TornadoLabs
    why: Offers curated video datasets and custom datasets for AI labs.
  • name: LAION
    website: https://laion.ai/blog/bvd/
    why: Its open video dataset is an alternative source for multimodal research.

keywords

  • multimodal training data
  • computer use datasets
  • video training data AI
  • expert gameplay dataset
  • speech dataset with transcripts
  • AI training data licensing

voice

  • technical and concise
  • scale-oriented
  • practical about modalities

signals

  • text: A lab is hiring for multimodal training data or computer-use research.
  • text: A team is launching a video, audio or agent model and needs expert demonstrations.

pricing

model: unknown

motion

hybrid

channels

  • self-serve platform
  • Hugging Face
  • direct sales

confidence

medium

Founders

  • name: Braiden Dishman
    title: Founder/CEO
    role: CEO
    bio: Co-Founder & CEO @ Shofo | UCSB Econ | Prev. AWS
  • name: Andre Braga
    title: Founder & Head of AI
    role: Founder
    bio: Co-Founder & Head of AI @ Shofo | UCSB BS Statistics & Data Science | Prev. MIT
  • name: Alexzendor Misra
    title: Founder/CTO
    role: CTO
    bio: Co-Founder & CTO @ Shofo | UCSB Dropout | Prev. CEO @ Correkt (43k users)

Back to YC companies