Phase 10: AI Engineering Career System · 45 min · Markdown · LinkedIn · Google Docs
The Concept
How Resumes Are Actually Read
Your resume goes through three screening passes before anyone reads it carefully. You must survive all three:
Pass 1 (machine, <1s): ATS parses text, matches keywords from the job description
Pass 2 (recruiter, ~7s): scans titles, companies, and the first bullet of each role
Pass 3 (hiring manager, ~1min): reads bullets for evidence of real, owned impact
Pass 1 — ATS (Applicant Tracking System): Software parses your resume, extracts text, and checks for keyword matches against the job description. If the job says "PyTorch, RAG, FastAPI" and your resume says "deep learning framework, retrieval system, web API," the ATS may not match you — even though you have the exact skills. The fix is not keyword stuffing; it's using the same terminology the job description uses, naturally, in context.
Pass 2 — Recruiter scan (~7 seconds): A recruiter reads 300 resumes a week. They scan vertically: job titles, company names, dates, and the first bullet of each role. If the first bullet doesn't demonstrate impact, they move on. They are not evaluating your technical depth — they are checking whether you look like someone who gets things done.
Pass 3 — Hiring manager (~1 minute): The hiring manager reads your bullets carefully, looking for evidence that you owned the work, not just participated. "Worked on" is a red flag — it suggests you were adjacent to the work. "Built," "led," "shipped" are green flags — they suggest you drove the outcome.
The Outcome-Driven Bullet
The unit that wins all three passes is the outcome-driven bullet:
action verb + what you built / scope + measurable outcome
WEAK: "Used Python and scikit-learn for a classification project."
STRONG: "Built a churn classifier (scikit-learn + FastAPI) serving 5k preds/day;
tuned the threshold to business cost, cutting missed churners 18%."
The weak bullet could describe a tutorial follow-along. The strong bullet tells the reviewer: what you built (churn classifier), the tech stack (scikit-learn + FastAPI), the scale (5k predictions/day), a specific engineering decision (tuned threshold to business cost), and the measurable impact (18% fewer missed churners). Every element does work.
Which resume bullet is stronger and why? A: "Used Python and scikit-learn for a classification project." B: "Built a churn classifier (scikit-learn + FastAPI) serving 5k predictions/day; tuned threshold to business cost, cutting missed churners 18%."
Bullet B follows the formula: action verb + scope/tech + measurable outcome. It tells the reviewer what you built (churn classifier), how (scikit-learn + FastAPI), the scale (5k/day), and the impact (18% fewer missed churners). Bullet A could be a tutorial follow-along — there's no evidence of ownership or impact.
What a Portfolio Project Needs
A GitHub repo with code is not a portfolio. A portfolio project is a communication artifact that lets a reviewer understand what you built, why it matters, and that you can ship — in 30 seconds:
The README is the most important file. It should answer: What problem does this solve? What does it do? What's the architecture? What were the results? A reviewer should understand your project from the README alone — without cloning the repo or reading the code.