Kimpton GTM brain

Kimpton builds custom AI evaluation and training environments for model labs and an AI research and trade-proposal system for investment teams.

category

AI evaluation and training environments

market profile

  • text: Koliseum creates synthetic reinforcement-learning environments from real-world data, with verifiable rewards and recorded trajectories.
  • text: Its investment product applies agents to filings, portfolios and market data while the investment team retains decisions.
  • text: Model evaluation buyers can also consider simulation and agent-testing platforms such as Patronus AI.

personas

  • title: Frontier model evaluation or reinforcement-learning lead
    pains:
    • Static benchmarks can be contaminated and fail to predict performance on genuinely new tasks
    • Building verifiable graders and detecting reward hacks or leaked state requires specialist effort
  • title: Fundamental equity portfolio manager
    pains:
    • Research across filings, transcripts and portfolio positions consumes time before a trade can be assessed
    • Needs evidence and risk context while retaining final investment authority

icp

  • text: Frontier AI labs and model teams evaluating agent behavior or training on domain-specific tasks with measurable outcomes.
  • text: Institutional fundamental equity managers with complex portfolios and existing market data; the initial target described by YC is $1 billion to $10 billion in assets.

message pillars

  • text: Custom evaluations and training environments address real tasks for frontier models.
  • text: Verifiable rewards make each failure cluster usable as a training environment.
  • text: Pinned evaluation versions enable comparisons across runs.
  • text: Investment professionals retain authority over agent-proposed trades.

objections

  • objection: Can an agent game the evaluation or see leaked state?
    response: Request the integrity test results and independently validate the graders and restrictions on representative tasks.
  • objection: Does the AI make trades without portfolio-manager oversight?
    response: Kimpton describes agents proposing trades with evidence, while the investment team makes decisions.
  • objection: Will evaluation results remain comparable as tests change?
    response: Ask how pinned versions, trajectories and execution settings are managed for your runs.

competitors

  • name: Patronus AI
    why: Offers simulation-based agent training and evaluation environments.
  • name: Arize AI
    why: Provides observability and evaluation for AI agents and applications.

influencers

  • name: Jack Zumwalt
    url: https://www.linkedin.com/in/jack-zumwalt
    why: Kimpton co-founder sharing product and benchmark updates.

keywords

  • AI agent evaluation environments
  • reinforcement learning environments
  • LLM benchmark contamination
  • verifiable reward environments
  • financial AI benchmark
  • AI investment research platform
  • model evaluation API

voice

  • Evidence-led
  • Technical and institutional
  • Direct and outcome-focused

signals

  • text: A model lab is training agents on specialized tasks and needs verifiable environments.
  • text: An evaluation team has reward-hacking or benchmark-contamination concerns.
  • text: A fund is exploring agent-assisted research with portfolio-manager review.

pricing

model: unknown

motion

hybrid

channels

  • developer documentation
  • LinkedIn
  • public benchmarks

confidence

medium

Founders

  • name: Jack Zumwalt
    title: Founder/CEO
    role: CEO
    bio: Co-founder, CEO @ Kimpton AI (YC P26). Building live evaluation arenas for financial work. Jack ran systematic strategies and managed daily operations of Level III Capital, a quantitative systematic trading firm he co-founded after raising $10M to seed the fund at the age of 22. Prior to that, he was a detection engineer at Goldman Sachs, writing algorithms to detect malicious threat actors who were trying to cyber attack the investment bank.
  • name: Adrian Del Bosque
    title: Founder/Founding Engineer
    role: Founder
    bio: Co-founder, Founding Engineer @ Kimpton AI (YC P26). Building live evaluation arenas for financial work. Adrian was a founding engineer at Level III Capital, a quantitative systematic trading firm, where he built the trading infrastructure. Prior to that, he was a Senior Cloud & Platform Engineer at Vistra, managing the cloud infrastructure behind the monitoring application for its 44,000 megawatt power generation fleet. He began his career at SpaceX as part of the Automation & Controls team.
  • name: Mauricio Ortiz
    title: Founder/CTO
    role: CTO
    bio: Co-founder, CEO @ Kimpton AI (YC P26). Building live evaluation readings for financial work. Mauricio oversaw the entire technical implementation of the investment process, from research to trade execution, at Level III Capital, a quantitative systematic trading firm he co-founded after raising $10M to seed the fund at the age of 21. Prior to that, he was part of the critical infrastructure team at Goldman Sachs which handled systems for its consumer bank and the Apple Card.

Back to YC companies