Riften GTM brain

Riften is an independent AI research lab working on model routing, evaluations and private model training for organizations building AI systems.

category

AI model infrastructure and research

market profile

  • text: Model routing addresses the tradeoff between task performance and the cost of different models.
  • text: Riften's public research identity is broader than its earlier gateway positioning; the current commercial offer requires confirmation.
  • text: Independent coverage describes a gateway and a longer-term goal of company-owned task-specific models.

personas

  • title: Head of AI infrastructure
    pains:
    • Inference costs rise when routine tasks use expensive frontier models.
    • Routing decisions need evaluation and fallback without rewriting applications.
  • title: Applied AI research lead
    pains:
    • Turning task outcomes into reliable evaluations and training data is difficult.
    • General models may underperform on recurring company-specific work.
  • title: Enterprise AI governance leader
    pains:
    • Needs clarity over who controls model weights and operational data.
    • Must assess quality and privacy before production traffic moves to a new provider.

icp

  • text: AI labs, AI-native companies and enterprises developing model-intensive applications, based on earlier public positioning; current customer profile is unconfirmed.
  • text: Likely early fit: engineering teams with repeated AI workloads, measurable task outcomes and enough inference spend to test routing; account size is unknown.

message pillars

  • text: Lead with independent research identity.
  • text: Connect research with useful systems.
  • text: Explore intelligence across digital and physical domains.

objections

  • objection: Will a lower-cost model degrade accuracy?
    response: Evaluate routing on representative tasks against a defined quality threshold and compare cost and fallback behavior.
  • objection: What data is retained and who owns trained models?
    response: Request current written data-handling, retention and ownership terms before sending sensitive workloads.
  • objection: How is this different from an existing gateway?
    response: Test whether the current offering combines routing with evaluation and private model work, and validate its production availability.

competitors

  • name: Martian
    why: Offers model routing and gateway infrastructure for multiple models.
  • name: Portkey
    website: https://github.com/Portkey-AI/gateway
    why: AI gateway offers model routing, fallbacks and governance.
  • name: OpenRouter
    why: Hosted API routes access to many language models.

influencers

  • name: Carl Stauffer
    url: https://mr6.foundation/found/riften-the-fight-for-the-prompt
    why: Founder and CEO associated with the company's model routing and ownership thesis.

keywords

  • LLM model routing
  • AI gateway
  • LLM inference cost optimization
  • AI model evaluation
  • private model training
  • open weight model deployment
  • physical AI research

voice

  • understated
  • research led
  • concise

signals

  • text: A team runs recurring model workloads and is scrutinizing inference cost versus quality.
  • text: A company wants to evaluate or develop private models from its own recurring tasks.

pricing

model: unknown

motion

unknown

confidence

low

Founders

  • name: Carl Stauffer
    title: Founder/CEO
    role: CEO
    bio: Founder and CEO of Riften. Former Palantir Forward Deployed Engineer. Former nuclear researcher at Jefferson Lab. CS and AMS at Johns Hopkins.

Back to YC companies