Lyon GTM brain

category

market profile

• text: Banks and fintechs use event histories across fraud, credit and customer models, often maintaining separate feature pipelines.
• text: Financial institutions must evaluate data isolation alongside predictive performance.

personas

• title: Head of credit risk modeling
  pains:

  • Default models miss temporal patterns across customer transactions
  • Feature engineering for each use case is slow
• title: VP data science at a fintech
  pains:

  • Separate fraud, churn and cross-sell models duplicate feature work
  • Needs measurable uplift over existing models
• title: Chief data or security officer
  pains:

  • Sensitive financial records cannot be freely sent to outside vendors
  • Model ownership and isolation require review

icp

• text: Banks, lenders, fintechs and payments businesses with large longitudinal transaction and customer event datasets.
• text: Organizations with cloud infrastructure and in-house data science teams able to evaluate downstream predictive performance.

message pillars

• text: Train a model on the institution's own event history.
• text: Use one representation across prediction tasks.
• text: Keep data and models within the customer's environment.

objections

• objection: Our existing tree models are accurate and explainable.
  response: Compare against the existing model on a time-separated holdout and evaluate governance and explainability before deployment.
• objection: Training a private model on sensitive data will be difficult to approve.
  response: Review the deployment architecture, security controls, ownership and costs with risk and cloud teams.

competitors

• name: Feedzai
  website: https://www.feedzai.com/resource/riskfm-foundation-model/
  why: RiskFM uses financial transaction data for risk modeling and fraud use cases.
• name: DataVisor
  website: https://www.datavisor.com/blog/top-10-real-time-decisioning-platforms-for-fraud-and-aml-in-2026
  why: Provides ML-based fraud and financial crime decisioning, an overlapping application budget.
• name: Oscilar
  why: Risk decisioning platform spanning credit and fraud.

influencers

• name: Markus Hoehn
  url: https://markushoehn.com/
  why: Lyon cofounder working on event data foundation models.

keywords

• transaction foundation model
• credit risk transaction data model
• private financial foundation model
• fraud prediction embeddings
• customer event data prediction
• banking model VPC

voice

• technical
• concise
• privacy-conscious

signals

• text: A lender is rebuilding credit risk models or seeking measurable lift in default prediction.
• text: A large fintech has multiple prediction projects on the same transaction history.

stack

• AWS
• Google Cloud
• Azure
• LoRA

pricing

motion

channels

• founder-led sales
• YC directory
• website meeting booking

confidence

Founders

• name: Gabriel Noya
  title: Founder/CEO
  role: CEO
  bio: Stanford EE and CS 26'
• name: Markus Höhn
  title: Founder
  role: Founder
  bio: Co-Founder at Lyon. Stanford Math