Pipeline overview

The Hermes resume pipeline is a staged process that narrows, enriches, and tailors information rather than trying to produce a final resume in one jump.

For the operator-facing walkthrough, use Resume Agent. This page is the compact internal pipeline reference.

This is the internal runtime stage flow from a queued JD to a pushed resume result.

• candidate-profile influences filtering, extraction, and tailoring.
• Pool evidence feeds project selection.
• The orchestrator handles the stage sequence.

Stage purposes

• resume-pipeline-orchestrator: fetch JDs, call downstream stages, and handle push and logging
• jd-prefilter: reject obvious bad fits quickly
• jd-extraction: turn raw JD text into structured downstream signals
• project-selection: choose supporting projects and OSS evidence only
• point-repointing: tailor selected projects plus all work experience without inventing facts
• latex-assembly: build the final resume artifact

Why this split works

Each stage reduces one failure mode:

• prefilter reduces wasted compute on bad roles
• extraction reduces vague JD interpretation
• selection reduces irrelevant project evidence
• repointing reduces generic resumes
• assembly reduces format drift

See also:

• Resume Agent > Skill Workflow
• Resume Agent > Run and Verify

If you want the actual skill files referenced in this section, use the public source repository.

hermes-autonomous-resume on GitHub