ATS engine
Layout structure, text layers, font safety, and table risk — the things that determine whether an applicant tracking system can extract the document reliably.
Analysis Methodology
Deterministic engines measure. AI explains.
ResuPulse analysis runs in two phases. First, seven deterministic engines measure your resume with plain code — no models, no sampling. Then a review layer of AI agents explains what the measurements mean and turns them into the guidance in your report. This page documents both phases so you know exactly which part of your report to trust as a measurement and which part is written advice.
Phase 1 — the seven deterministic engines
Each engine measures one aspect of the document. They run in parallel and produce structured measurements that the rest of the pipeline never reinterprets.
Structure & formatting engine
Margins, typography consistency, and content that spills beyond a page — presentation problems that survive extraction.
Impact engine
Action verbs and the ratio of quantified achievements — how much of your experience is written as outcomes rather than duties.
Skills engine
Skills taxonomy coverage and evidence mapping — whether the skills you list are connected to the work that demonstrates them.
Content engine
Summary signals and duplicate bullets — whether sections repeat themselves and whether your summary carries real signal.
Experience engine
Sweep-line tenure calculation, career gaps, and progression — your timeline measured consistently across the whole resume.
Projects engine
Project complexity tiers and link checks — whether projects are described with enough structure to be evaluated.
Phase 2 — the AI review layer
After the engines finish, AI agents write the report. A positioning agent runs first and establishes your candidate context — target role, career level, primary domain, and the deterministic total years of experience. Specialist agents then review the same measurements in parallel: ATS readiness, content and impact, skills gaps, writing diagnostics, experience depth, projects, a recruiter-take summary, and the action plan.
The division of labor is deliberate: numbers are the engines' job, and the words are the agents' job. The AI layer is never allowed to recalculate anything numeric.
What is guaranteed
- Engine measurements are deterministic: the same resume produces the same numbers every time.
- The AI layer never recalculates dates, total experience years, word counts, or finding codes — those always come from the engines.
- Your total years of experience is always pinned to the deterministic calculation, never to a model's estimate.
- In production there are no fallback or mocked report values: if a step fails, it fails loudly rather than shipping made-up data.
What the analysis cannot do
- The analysis evaluates presentation and structure, not truth. It cannot verify that a bullet point is accurate.
- The AI review produces guidance to edit — it never edits for you or applies changes automatically.
- Scores measure readability and clarity in ResuPulse's framework. They are not predictions of hiring decisions.
The scoring dimensions and weights are documented separately on the ATS score methodology page.
Frequently asked questions
What is the difference between the engines and the AI review?
The engines are deterministic code that measures the document — extraction, layout, impact verbs, skills coverage, tenure, and project structure. The AI review then interprets those measurements and writes the guidance you read. The numbers come from code; the explanations come from the AI.
Why does the same resume always produce the same engine results?
Because the measurement layer contains no randomness. AI models are probabilistic, so the analysis pipeline keeps them out of anything numeric — scores, dates, counts, and finding codes are always produced by the deterministic engines.
Does the AI layer ever change my resume?
No. The AI review explains findings and suggests edits. Any change to your resume is made by you, in your own document.
Where do the score weights come from?
The scoring dimensions and their weights are published on the ATS score methodology page.