Founding Engineer (Full-Stack)

About Harness
Harness is building the AI-native commercial insurance brokerage.
Commercial insurance is a $1T+ market, but the brokerage layer still runs on manual judgment, fragmented workflows, repetitive coordination, and tribal knowledge. A great broker does much more than sell insurance. They understand a business, identify its risks, determine which insurance companies are likely to write it, prepare the right submission, compare quotes, guide the customer, and get the policy bound.
Today, nearly every step of that process requires humans to move information between systems, coordinate across different parties, and decide what happens next.
Harness is building software to automate that entire funnel.
We break the brokerage workflow into its underlying decisions and tasks, identify the bottlenecks, and build systems that make each step faster, smarter, and increasingly autonomous. The goal is not to build a better CRM for insurance brokers. The goal is to build an autonomous brokerage.
Harness started in January 2026. In our first 8 months, we grew to $1M ARR and are targeting $3M ARR by the end of 2026.
We are well-funded, growing quickly, and still early enough that the systems you build will define how the company operates.
What We Are Building
Think about the lifecycle of a commercial insurance account:
Lead → Qualification → Intake → Risk Understanding → Coverage Requirements → Market Selection → Submission → Carrier Follow-Up → Quotes → Comparison → Negotiation → Binding → Servicing
At every stage there are decisions, dependencies, and operational bottlenecks. Our job is to understand those workflows deeply and progressively automate them.
Our engineering process is straightforward:
Understand how the workflow works today
Identify the biggest bottleneck or source of manual work
Determine what judgment, context, or data is required to complete it
Build software or AI systems that can perform that work reliably
Measure the impact
Move to the next bottleneck
Over time, individual automations become connected systems, and connected systems become an increasingly autonomous brokerage.
That means we work on problems like:
Turning conversations, documents, and external data into structured risk information
Determining what information is missing from an account
Selecting the right insurance markets based on appetite and underwriting criteria
Generating and improving carrier submissions
Automating communication and follow-ups with customers and underwriters
Tracking asynchronous workflows that may run for days or weeks
Building agents that know when to act, when to wait, and when to escalate
Extracting knowledge from successful and unsuccessful placements
Designing reliable human-in-the-loop systems for high-stakes decisions
Building the infrastructure that allows hundreds or thousands of accounts to move through the brokerage simultaneously
The Role
We are hiring engineers to build the core systems behind Harness. This is primarily a backend and systems engineering role, but we do not have a hard boundary between backend and frontend.
You will build the infrastructure, data models, APIs, workflow engines, agent systems, integrations, and product surfaces that power the brokerage. If you build a system, you should also be comfortable taking it all the way to the user when necessary, including building the interface that makes it usable.
You will work directly with the founders. You will not receive polished specifications describing exactly what to implement.
Instead, you might hear:
“We lose too much time waiting for this information.”
or
“Our brokers have to manually decide where to send every account.”
Your job is to understand why, study the workflow, determine what can be automated, design the system, ship it, and measure whether it actually improved the business.
The best engineers here will spend as much time understanding the problem as writing the code.
What You Might Build
Depending on the week, you might:
Design a workflow engine coordinating long-running insurance processes
Build an AI agent that determines the next action required on an account
Create systems for extracting structured insurance data from documents and conversations
Model complex commercial insurance accounts and their relationships
Build carrier appetite and market-selection infrastructure
Integrate with carrier portals, email, telephony, document systems, and third-party data providers
Build evaluation and observability infrastructure for production AI agents
Create systems that automatically detect stalled accounts and unblock them
Design APIs and event-driven systems connecting different parts of the brokerage
Build interfaces for brokers to supervise and correct AI actions
Take a workflow that currently takes a broker 30 minutes and reduce it to 30 seconds, or eliminate it entirely
What We’re Looking For
2+ years of strong backend engineering experience
Experience designing APIs, services, data models, and distributed systems
Strong proficiency with relational databases and data modeling
Strong Python, TypeScript, Go, Java, or equivalent backend experience
Experience building production AI agents, LLM-powered systems, or workflow automation
Understanding of the reliability challenges involved in putting LLMs into real production workflows
Ability to decide when to use an LLM, when to use deterministic software, and how to combine the two cleanly
Strong debugging and systems thinking
Ability to simplify complicated business processes into clean abstractions
Heavy use of modern AI development tools such as Claude Code, Cursor, ChatGPT, Codex, or similar
High ownership and agency
How We Think About Engineering
We care much more about business leverage than lines of code.
A great engineering project might involve building an entirely new system. It might also involve discovering that a workflow requiring five people and twelve steps can be redesigned into two steps and automated with 500 lines of code.
We expect engineers to ask:
Why does this process exist?
What decision is actually being made?
What information is required to make it?
Can software make that decision?
Can AI make it reliably?
What happens when it is wrong?
How do we measure whether this made the brokerage better?
We use AI heavily in engineering ourselves. Tools like Claude Code, Cursor, ChatGPT, and Codex substantially increase how much a small engineering team can ship.
We therefore optimize for engineers who can own large problems rather than narrowly scoped tickets.
Why Join Now
Harness has already proven that the business works. Now the challenge is fundamentally different.
We need to turn what our team does manually today into software and AI systems that can operate the brokerage at dramatically greater scale. That means systematically finding bottlenecks across the entire insurance funnel, understanding the judgment and coordination behind each one, and automating them one by one.
The goal is for the same team to eventually handle 10x the volume.
This is where engineering becomes one of the primary constraints on growth. The systems you build will directly determine:
How many customers a broker can manage
How quickly an account moves from intake to binding
How intelligently we select and approach insurance markets
How much customer and carrier communication can happen automatically
How much human coordination can be eliminated
How quickly Harness can grow without adding people at the same rate
We already have the customers, real workflows, operating data, and a brokerage team showing us exactly where the bottlenecks are. The next phase is about encoding that knowledge into software and building systems that let the business scale without scaling human effort at the same rate.

Work arrangement
No