Perceptual Labs GTM brain

category

market profile

• text: Businesses with domain-specific data may need specialized models when general models miss their cost, latency or quality requirements.
• text: The company's Foundry describes a training stack spanning data generation, reinforcement learning environments and continuous learning.

personas

• title: Head of AI or machine learning
  pains:

  • General models underperform on company-specific data or specialist tasks.
  • Building training infrastructure and evaluating model improvements requires scarce expertise.
• title: CTO or VP engineering
  pains:

  • High inference cost and latency constrain production deployment.
  • Wants control of model weights and data rather than relying exclusively on a shared model.
• title: Search or document automation product lead
  pains:

  • Generic retrieval misses catalog terminology and domain-specific meaning.
  • Document extraction must handle variable scans and forms at production scale.

icp

• text: Businesses with proprietary documents, catalogs, images or other domain data and a repeatable model use case.
• text: Engineering-led organizations with enough task volume to care about model ownership, latency and inference economics.
• text: Teams developing search, document classification or multimodal products that can provide domain expertise and evaluation data.

message pillars

• text: Customers retain control of their models and data.
• text: Specialize intelligence to the business and its data.
• text: Build across modalities beyond text.

objections

• objection: Why not use a general model API?
  response: Compare quality, latency and cost for a representative domain task; Perceptual argues specialized models can improve those tradeoffs.
• objection: Will we own what is trained?
  response: The company positions the offering around customer ownership of weights and data; confirm contract and base-model license terms.
• objection: We lack a ready training dataset.
  response: Foundry describes synthetic and curated domain data generation; assess data availability and quality in discovery.

competitors

• name: Together AI
  website: https://www.together.ai/custom-training
  why: Offers custom training and reinforcement learning for models.
• name: Amazon Web Services
  website: https://aws.amazon.com/sagemaker/ai/model-customization/
  why: SageMaker enables model customization and customer control over open model weights.
• name: POFE
  website: https://pofe.ai/
  why: Provides model post-training, evaluation and downloadable weights.

keywords

• custom AI model training
• enterprise model fine tuning
• own AI model weights
• multimodal model training
• custom embedding model
• document classification model
• AI inference cost reduction

voice

• Ambitious and ownership-oriented
• Technical yet concise
• Focused on business outcomes

signals

• text: A team has rising inference costs or latency for a repeated AI task.
• text: A business is building search over a specialized catalog or processing high volumes of documents.
• text: An AI leader seeks control over model weights and proprietary training data.

pricing

motion

channels

• website contact
• X
• research content

confidence

Founders

• name: Kevin Miao
  title: Founder/CEO
  role: CEO
  bio: Co-founder at Perceptual Labs. Senior Research Engineer at Apple and Lecturer at UC Berkeley. Previously researcher at Berkeley AI Research (BAIR). MS and BA in Computer Science from UC Berkeley. Born and raised in the Netherlands; speak 8 languages!
• name: Frankie Li
  title: Co-Founder
  role: Founder
  bio: Co-Founder at Perpetual Labs. Previously helped train models and led product at Fundamental Research Labs (a16z-backed, MIT spinout), hire #3 where he scaled products to millions of users and revenue, including Shortcut (Excel copilot used by 1000+ companies). Harvard Grad.