Intern, AI Agent Evaluations

G2•San Francisco, CA
•$24 - $24

About The Position

The Intern, AI Agent Evaluations supports G2’s Agentic Marketplace team in expanding its program for evaluating AI agents on real tasks and publishing the results as leaderboards on g2.ai. The role contributes to evaluation methods, vendor engagement, and market monitoring. It reports to the Head of Product, Agentic Marketplace, and works with product, data science, and engineering. About half the role focuses on evaluation methods: breaking an AI agent’s job, such as outbound sales, into testable steps; writing tasks, test data, and scoring rules; and reviewing scores from AI judges and human reviewers. About a quarter involves vendor relationships, including supporting a system for engaging hundreds of vendors, preparing results packets before publication, and keeping the vendor pipeline current weekly. The remaining time focuses on tracking other AI agent evaluators and the results they publish.

Requirements

  • Currently enrolled in an MBA program.
  • Have built with or used AI tools such as LLM APIs, chat assistants, or agents, and be able to explain how they work and where they fail.
  • Write clearly and briefly.
  • Be comfortable speaking with startup founders and product leaders, holding a view when challenged, and changing it when facts warrant.
  • Be able to make progress on problems without an established process.
  • Be based in the San Francisco Bay Area.
  • Be available 10 to 20 hours per week during the school year.

Nice To Haves

  • Have built an AI agent, evaluation, or benchmark, including as a class or side project.
  • Have worked at or with a B2B software company in any role, such as startup, venture, sales, or product.
  • Have worked in customer support, sales, or marketing.
  • Have written about AI or software for a public audience, such as through a blog, newsletter, or research.
  • Already know the AI startup scene in San Francisco.

Responsibilities

  • Break AI agents’ jobs into steps that can be tested.
  • Write evaluation tasks, test data, and scoring rules.
  • Review scores assigned by AI judges and human reviewers.
  • Support a system for engaging with hundreds of vendors.
  • Prepare results packets to share with vendors before results are published.
  • Keep the vendor pipeline current each week.
  • Track organizations that evaluate AI agents, including AI labs, benchmark groups, evaluation startups, and analyst firms, and monitor what they publish.
© 2026 Teal Labs, Inc
Privacy PolicyTerms of Service