Harper is an AI-native commercial insurance company in San Francisco. We're not bolting AI onto insurance — we're rebuilding the entire business as software, on a simple bet: turning expert human judgment into compute is one of the largest transitions left to make, and a trillion-dollar industry still run 90% by hand is the place to prove it. We've grown ~100x in the last year and we move at that speed — on-site, in person, long days, very high standards. Almost no one joins Harper for insurance; they join to build the company that replaces how it works.
The role
The model is only one part of a worker. The loop, the tools, the memory, the execution environment, the guardrails, the fallback path, the observability — that's what decides whether an agent can actually do useful work inside a real company. That's the harness, and at Harper every agent (product and coding) rides on a shared one. If the harness is brittle, every pod ships brittle agents; if it's good, pod engineers stop rebuilding the same scaffolding and start shipping agents in days instead of weeks. That makes the harness the company-specific leverage behind the whole bet — the model is replaceable, the substrate beneath it is the moat. You'll build inside the harness layer — the meta-harness wrapping our frontier coding agents, OpenClaw, Hermes, and our internal agents — working alongside the engineer setting harness direction, shipping the primitives every agent team depends on.
What you'll do
Build agent-loop primitives. Prompt construction, tool routing, retry/timeout/budget logic, multi-step orchestration.
Ship execution-environment infra. Sandbox lifecycle, isolation, blast-radius limits, filesystem and network policy for agents.
Own the tool layer for your domains. Schema, auth, rate-limit, observability per tool — tools other engineers build on.
Wire model-provider routing. Provider fallback chains, eval-driven model selection, cost/latency tradeoffs.
Ship harness SDK improvements — whatever makes pod engineers faster.
Eat your own dog food. You write code with our harness daily and feel every rough edge before anyone else reports it.
What we're looking for
3–6 years building software, with production agentic systems behind you (real users, real traffic — not demos) and experience with at least one major agent framework.
You think in tool design, not just prompt design, and can describe a tool-design or sandbox decision you'd still defend in three years.
Production agent/LLM systems experience: agent loops, tool integration, prompt engineering at scale.
You write code with AI daily and manage 3+ parallel sessions.
Strong written communication (API contracts, integration guides, internal docs).
Bonus: sandbox/isolation infrastructure; open-source agent-framework contributions; foundation-model partner or early-access experience.
The reality
On-site in San Francisco, in person, long days, high standards. The harness is infrastructure other engineers build their work on, so the bar is high and the blast radius is real — when you get it right, the whole org's agent velocity moves; when you don't, everyone feels it. The right person wants that leverage and that pace.
Logistics
Compensation (OTE): $187,000–$264,000 cash (base + target performance bonus), plus competitive equity.
Location: San Francisco, in-office. Based here or willing to relocate.
Benefits: Uber commuter benefits; breakfast, lunch, and dinner provided; snacks and coffee stocked; free gym membership; health, dental, and vision.
Process: Founder call (15 min) → Tech Lead deep-dive (60 min, harness) → Super Day on-site → founder + Tech Lead offer. No committee. Best offer, first.
To apply: If you've shipped an agent loop or a tool layer that real engineers depend on — send your resume, the code, and tell us about a harness decision you'd defend.