Member of Technical Staff, Agent Platform

Perimeter Compute
•$150,000 - $250,000•Remote

About The Position

Perimeter Compute is hiring a Member of Technical Staff to own the intelligence behind its agents: the models they use, the knowledge they rely on, and the skills and tools they work with. The role covers every agent the company runs, from coding agents to agents embedded in its products. A central focus is ensuring that every agent action is attributable to the person whose intent and authority it carries, while keeping the work agents hand to people worth their attention. The successful candidate will measure progress by how often agents complete tasks without human intervention, how well that work holds up 30 days later, and the cost of each completed task. Agents already ship hundreds of changes a day, giving the role rapid feedback on whether improvements work. This is a fully remote role in the United States. Perimeter Compute is unable to sponsor visas at this time. Applications are accepted on an ongoing basis.

Requirements

  • Have built and operated production software that people depend on, and taken ownership when it broke.
  • Have shipped an agent or LLM system that other people relied on every day.
  • Run coding agents on real work and have strong opinions about where they fail.
  • Be able to show whether a change to an agent made it better or merely different.
  • Have built guardrails that people did not route around.
  • Have written instructions or skills that agents follow reliably.

Nice To Haves

  • Experience running Claude Code, Codex, or similar agents across a whole team.
  • Experience with evaluations for agents or LLM systems.
  • Experience with identity, permissions, or audit systems that track who did what on whose behalf, such as OAuth token exchange.
  • Experience with tool integrations using MCP or similar.
  • Experience building sandboxes or throwaway environments that start in seconds.

Responsibilities

  • Choose and route between models, and keep agents working as models change.
  • Own the shared knowledge every agent starts from, with a named owner for each fact and a way to retire stale information.
  • Own the standard for agent skills, and test every change to a skill, prompt, or model against real tasks from the company's history before it ships.
  • Build fast, throwaway sandboxes where agents can work without touching anything shared.
  • Make every agent action trace back to the person who requested it, with only the authority they gave it, working closely with the security team.
  • Identify where agents fail or waste people's attention, and incorporate fixes into the system.
  • Turn what works in launches into building blocks the rest of the team can use to put reliable agents into products.

Benefits

  • Health, dental, and vision insurance.
  • Unlimited PTO.
  • A $500 home office stipend.
  • A top-spec MacBook.
  • Unlimited tokens.
© 2026 Teal Labs, Inc
Privacy PolicyTerms of Service