Carpe

Job Description About Carpe & the Role Carpe is an AI-native company building platforms and data products that help P&C insurance carriers make better decisions. At Carpe, we don't hand-write code anymore. We build with AI. As a Senior AI Software Engineer, you will turn business needs and prototypes into reliable production systems by directing frontier models and coding agents, shaping the environments they work in, and verifying what they deliver. AI is your primary way of building-not an autocomplete tool added to a traditional coding workflow. You bring extensive hands-on experience with frontier models, agents, and agent harnesses, plus the engineering judgment to take their work all the way through AWS deployment and production operation. You'll have a say in how we build. Our design and architecture patterns give you a solid starting point, not a set of limits. If a new approach or technology will better serve our customers, make the case. We care far more about how you think through tradeoffs than whether you've used every tool we use. Responsibilities Turn business needs into executable work. Partner with Product to define the problem, constraints, acceptance criteria, and customer outcome. Give agents the specifications, repository context, examples, and verification steps they need to deliver the right solution; resolve ambiguity before it becomes working software that solves the wrong problem. Architect systems and direct AI delivery. Break complex work into bounded tasks, choose interfaces and data flows, and direct agents through implementation, integration, and debugging. Use our approved patterns to address scale, latency, reliability, and data correctness across products. You own the architecture and outcome, even when AI produces the code. Build effective agent workflows and harnesses. Scope and steer bots through our team workflow, including Slack and Linear. Configure the context, tools, permissions, execution environments, and feedback loops around agents; manage handoffs, checkpoints, and recovery when they stall or go off course. Run a tight build-and-verify loop: the agent builds, you exercise the result in a live test environment, feed back precise findings, and repeat until the acceptance criteria are met. Choose models and approaches deliberately. Evaluate frontier models and coding agents on real engineering tasks, comparing output quality, reliability, context handling, latency, and cost. Match the model and harness to the work, and adapt as capabilities change. Let customer outcomes, not tool loyalty, guide decisions. Prove the result, not the agent's claim. Review architecture and generated changes at the level the risk demands; personally inspect sensitive changes. Use independent checks, representative data, automated tests, regression checks, and end-to-end behavior to catch plausible-but-wrong output. Where systems use models at runtime, evaluate their behavior and failure modes too. Ship and operate in AWS. Take AI-built software through cloud deployment and production support with QA and DevOps. Direct and verify infrastructure-as-code and CI/CD changes; understand access controls, secrets, networking, data storage, rollout, and rollback. Use logs, metrics, and traces to diagnose failures, verify recovery, and own incident follow-through-including failures introduced by agents. Make the next AI build better. Turn successful approaches and failures into reusable agent instructions, repository guidance, harness improvements, and verification routines. Keep architecture and operational context current so both people and agents can build on it. What we are looking for 7+ years of professional software engineering, delivering and supporting production systems (overall experience, not five years with today's AI tools). Extensive frontier-model experience. Substantial use of frontier models across multiple model families-for example, OpenAI, Anthropic Claude, and Google Gemini-in real software delivery. You can explain which models you choose for planning, implementation, debugging, and review; where they fail; and how you evaluate a new release. You actively experiment and clearly prefer building with AI over a hand-coding-first workflow. Extensive agent and harness experience . You have shipped meaningful work with coding agents and agent harnesses-not just chat prompts or autocomplete. Examples may include Claude Code, Codex, other agentic development environments, or custom harnesses. Extra credit for OpenClaw. You can configure repository instructions, context and tool access, isolated execution, task decomposition, verification, and recovery, and explain when orchestration helps versus adding overhead. End-to-end systems ownership. You have personally designed a complete system through professional or substantial independent work, with relevant professional experience around similar systems. Show how you directed AI from an initial specification through a working, deployed result, what you personally decided, and how you established that it worked. You don't need to have led a large-system initiative. Engineering depth behind the agents. Strong programming, debugging, architecture, and systems reasoning. You can read and interrogate generated code, trace data flows and service interactions, identify flawed assumptions, and steer a correction. We value your ability to make AI-built systems correct and maintainable-not the volume of code you type. AWS and cloud deployment experience. You have deployed and supported production software in AWS and understand the path from repository to running service: infrastructure as code, CI/CD, compute or containers, storage, networking, IAM, secrets, observability, and rollback. Explain a deployment or incident you owned and how you verified the outcome. Roll-up-your-sleeves, low-ego collaboration . Comfortable steering agents, investigating failures, explaining tradeoffs, and influencing without direct authority in a small. Share work and blockers openly, move across products, and help teammates improve their AI workflows while staying accountable for the customer outcome. Why Carpe? Your work shows up fast. At our size, what you build reaches customers quickly, and you quickly make an impact Growth: We have a clear growth framework with defined levels and competencies, so you always know what's expected and what "next" looks like. Collaboration: Work with teams in sales, marketing, engineering, product, and customer success. We Live our Values: Be Bold, Be Creative - Lead with Empathy - Act Like an Owner Flexible working hours and flexible vacation policy 100% Carpe-paid health, dental, and life insurance (Additional benefits at Employee cost) Full-time permanent positions are eligible for the Carpe Data Bonus Plan. 401k with company matching Location We are headquartered in Santa Barbara, CA. We work on a hybrid schedule: Tuesday/Wednesday/Thursday in the office. Salary Range $155,000 - 180,000. The actual compensation for this role will be determined by a variety of factors, including but not limited to the candidate's skills, education, experience, and internal equity.