Software Engineer - AI Automation

Coalition is the world's first Active Insurance provider designed to help prevent digital risk before it strikes. Founded in 2017, Coalition combines comprehensive insurance coverage and innovative cybersecurity tools to help businesses manage and mitigate potential cyberattacks.

Opportunities to make an impact with bold thinking are real—and happening daily at Coalition.

We are hiring a Software Engineer for the Automation team to use AI to solve and automate manual work across Coalition. This is a hands-on AI engineering role: authoring skills, wiring agents to our systems of record through MCP, building evals, and deploying agents that non-engineers depend on every day. You will not be starting from zero. We have early workflows live in production, more proofs of concept in the oven, and working hypotheses about how all of this should fit together. We want someone who will pressure test those hypotheses against their own experience and change our minds where we have it wrong. Deep, current, practical experience building with LLMs is the single most important qualification for this job.

The other half of the role is proximity to the business. Think of this as a Forward Deployed Engineer, pointed inward: your customers are other Coalition teams. You will work directly with the teams whose work we are automating, learn how they actually operate, and turn that into a playbook. Getting from that conversation to something an agent can run reliably is the core of the work. You will also work day to day with our data team to get at the data these agents depend on. This is an individual contributor role with no direct reports.

• Author skills: Turn a team's process, judgment, and edge cases into reusable skills that agents can invoke. Write them so the next project can reuse them instead of rebuilding them.

• Build within our agent architecture: Wire agents to the systems of record through MCP servers, compose skills and tools into working workflows, and add new MCP surface area when a system we need is not covered yet.

• Build evals: Define what "correct" means with the business team before you build. Build the eval set, establish a baseline, and use it to make calls about what ships. Nothing goes live on vibes.

• Deploy and support agents: Take agents from prototype to production, including permissions, human-in-the-loop checkpoints, audit trails, and rollback. Tune from how they behave once real users are in front of them, and leave behind docs and an owner on the business team who can run it without you.

• Partner with the data team: Work with our data engineers on the access, models, and quality of the data your agents read and write. Bring them requirements early rather than working around them.

• AI application building (primary): Substantial hands-on experience shipping LLM-backed applications: prompting, tool use, agentic loops, and multi-step workflows. You know where these break and how to contain the damage when they do. You have shipped at least one agent that non-engineers depend on.

• Agent tooling: Practical experience with MCP, skill or tool authoring, and orchestrating agents against real systems of record.

• Evals: You have built eval sets and used them to make decisions, not just to make a slide look good.

• Coding agents: Fluent with agentic coding tools like Cursor. You use them to move faster inside codebases you did not