Founding Harness Engineer

Heyjared; San Francisco (CA); On-site; USD 250,000 - 350,000; Full time

Benefits offered by this job

Equity

Job summary

Heyjared in San Francisco invites a founding Harness Engineer to build an agent runtime that acts as a proactive yet prudent teammate. You’ll design prompts, implement tool use, and ensure safe, multi-tenant data practices as you collaborate with the founder and early users.

You will ship production-grade agents, craft robust evaluation harnesses, and drive prototypes to production, with equity and a high-impact founding role in a hands-on, in-person setting.

Qualifications

• Shipped agents that use tools in production, not demos or prompt wrappers.
• Hands-on system prompt design improved through testing.
• Strong experience with regression/eval harnesses.
• Knowledge of safe tool execution and data risk management.

Responsibilities

• Design and build the agent runtime: how Jared gets context, decides when to speak, chooses tools, waits for approval, completes work.
• Write and iterate system prompts: proactive without noisy or reckless.
• Build eval and regression tests around real tasks, tool use, failure cases.
• Make behavior reliable across long-running work, interruptions, retries.
• Build safe boundaries: approval gates, multi-tenant data separation.
• Work directly with the founder and early users.

Skills

Agent orchestration

System prompts

Regression harness

Tool integration

Security boundaries

Job description

Founding Harness Engineer - San Francisco

Jared - Hire an A[I] Player

Location: San Francisco, in person. No remote. Bay Area or willing to relocate.

Compensation: $250,000-$350,000 base salary plus equity.

Equity percentage, vesting and (where applicable) commission plan require confirmation. OTE is not guaranteed base pay.

Jared is an AI coworker for teams. You'll build the harness that makes it act like a useful teammate: proactive when it should be, careful when an action needs approval, dependable enough for real work.

What you'll do:

• Design and build the agent runtime: how Jared gets context, decides when to speak, chooses tools, waits for approval, completes work
• Write and iterate system prompts: proactive without noisy or reckless
• Build eval and regression tests around real tasks, tool use, failure cases. Prompt changes are code: test them, catch regressions
• Make behavior reliable across long-running work, interruptions, retries
• Build safe boundaries: approval gates, multi-tenant data separation
• Work directly with the founder and early users

What we're looking for:

• Shipped agents that use tools in production, not demos or prompt wrappers
• Knows how to get an agent to take useful initiative, and when it should act, ask, or stay quiet
• Hands-on system prompt design improved through testing
• Strong regression/eval harness experience
• Understands safe tool execution, approval gates, multi-user data risk
• Owns broad problems prototype to production, explains tradeoffs
• Wants a founding role, in person in SF