AI Agent Engineer — Early Career

CaelumenAI is building the governed production layer for agent-native software. PERCO, our flagship product, turns agent output into a reviewable plan, a controlled release, and evidence of what happened.

About the role

As an AI Agent Engineer — Early Career, you will own small, real product slices across agent workflows, developer tools, APIs, browser experiences, and deployment verification. You will work with coding agents as collaborators, review their output, trace failures to root causes, and carry changes from a clearly scoped problem through implementation and real-world evidence. This role is open to new graduates and other early-career engineers with roughly zero to two years of professional experience. Strong project evidence can come from coursework, open source, internships, personal products, or paid work.

Applicants must have valid authorization to work in the United States when employment begins. We welcome candidates working under valid OPT and support eligible H-1B sponsorship.

Responsibilities

• Build agent-native product and developer workflows across Python, JavaScript or TypeScript, APIs, CLI, MCP, and browser surfaces
• Turn product problems into small end-to-end implementations with clear acceptance evidence
• Use coding agents effectively by preparing context, reviewing output, and retaining human ownership of consequential decisions
• Create automated coverage with pytest, Playwright, integration tests, and agent evaluations
• Investigate failures to root cause and ship durable fixes with regression coverage
• Improve the reliability, clarity, and safety of PERCO's human-controlled release workflows

Requirements

• Evidence over pedigree: show us something you scoped, built, tested, and debugged—and tell us what the evidence changed
• Approximately zero to two years of professional software engineering experience, or equivalent evidence through substantial projects
• Working proficiency in Python, JavaScript, or TypeScript, plus Git and web or API fundamentals
• One project you scoped, built, tested, and debugged that you can explain in detail
• Hands-on experience using or evaluating coding agents in a real development workflow
• Comfort reading logs, testing real user journeys, and iterating until a system works end to end
• Clear written communication, careful judgment, and a bias toward ownership

Skills

• Experience with MCP, tool schemas, agent frameworks, or agent evaluation
• Experience building CLIs, developer tools, or browser-based products
• Familiarity with cloud services, containers, CI/CD, or infrastructure as code
• Open-source contributions or a deployed product used by other people

Benefits

• Direct mentorship on real agent-native product work
• High ownership over scoped, end-to-end engineering projects
• Flexible remote collaboration with an option to work from San Francisco
• Work across the complete loop from agent interaction through verified production delivery
• Support for candidates working under valid OPT and eligible H-1B sponsorship