AI Engineer Dojo Contents
AI Agent Engineer · Career

Landing Your First Agent Role

You've learned the loop and built a harness. Here's how to turn that into a job.

Build a portfolio — this is the whole game for new grads

Take the agent you built and point it at a real, self-contained task: a coding agent that fixes a failing test in a small repo, a research agent that answers a question by calling a search tool and citing sources, or a data agent that queries a SQLite DB to answer questions. Give it real tools, a stopping rule, and a permission gate on anything destructive. Then wrap it in an eval harness that runs it over a dozen tasks and reports task-completion rate, average steps, and cost. Put it on GitHub with a short write-up: the task, where the loop failed (looping, wrong tool, never stopping), what you fixed, and the before/after numbers. One real, measured agent beats a page of buzzwords — and almost no other new grad will have one. (Marcus, in Chapter 12, got hired on exactly this.)

What to search for

Titles vary: AI Agent Engineer, Applied AI Engineer, AI Software Engineer, LLM Engineer, Forward-Deployed Engineer, Agent / Automation Engineer. Also look inside broad "AI Engineer" and "Applied ML" postings — building and hardening agent loops is often the bulk of the work even when it isn't in the title. Anything described as "AI that takes actions" or "autonomous workflows" is an agent job.

What entry-level interviews reward

Systems thinking over jargon. Handed "our agent sometimes loops forever / costs too much / does the wrong thing," they want to hear you reason about the loop — stopping rules, tool design, step and cost budgets, retries, and permission gates — and, above all, how you'd measure completion and failure over many runs rather than eyeballing one. The Interview Simulation appendix drills exactly this — practice it out loud.

Resume framing

Lead with what you built and measured, not courses taken: "Built a tool-using coding agent with a step budget and permission gates; evaluated it over 30 tasks, raising task-completion from 55% to 80% while cutting average steps from 14 to 8." Numbers and verbs — that one line signals you can do the job.

A two-week plan

Week 1: read Ch 1–6, do every lab. Week 2: read Ch 7–12, do every lab, then build the portfolio agent on a real task. Rehearse the ten interview questions below. Apply.


← Back Continue →

AI Agent Engineer — New-Grad Edition · AI Engineer Dojo · aiengineerdojo.com