Resume Bullet Writer

Purpose

Turn verified experience/project evidence into readable, interview-defensible bullets.

A strong bullet does not need a number. A precise technical implementation is stronger than an invented metric.

Bullet Patterns

Direct Implementation

"Implemented RBAC with JWT and Express middleware across candidate and recruiter routes."

Architecture / Technical Decision

"Built a FastAPI diagnostic pipeline combining integrity checks, drift detection, SHAP feature analysis, and historical-case retrieval."

Measured Result

Use only with verified provenance: "Reduced median search latency from 310 ms to 140 ms in a 1,000-query local benchmark by adding compound indexes."

Research / Evaluation

"Achieved 0.84 macro F1 on the held-out test split using [model], evaluated on [dataset/scope]."

PACTI — Optional

Use selectively:

• Problem
• Action
• Core technical decision
• Technical implementation
• Impact

Do not force all five parts into every bullet.

Metric Rules

Use a number only when supported by:

• benchmark/test output;
• analytics/logs;
• evaluation report;
• repository-visible stable count;
• documented work record;
• candidate-provided measurement that can be explained.

For every metric ask:

• How was it measured?
• In what environment?
• On what sample or time period?
• Can the candidate reproduce or defend it?

If those answers are unavailable, remove the number.

Impact Context

• Production/operational — only for real live operational measurements.
• Research/evaluation — state dataset/split/metric context.
• Development benchmark — explicitly say local/testing/development.
• Descriptive — implementation detail with no fabricated outcome.

Writing Rules

• Start with a specific action.
• Name the system/technology when relevant.
• Keep ownership accurate.
• Prefer one core idea per bullet.
• Avoid vague filler.
• Avoid inflated verbs when a plain verb is more accurate.
• Do not repeat keywords for density.
• Keep bullets short enough to scan normally.

Before / After

Weak: "Worked on authentication."

Better: "Implemented JWT authentication and role-based authorization middleware for candidate and recruiter routes."

Weak: "Improved database performance by 70%."

Better when no benchmark exists: "Added MongoDB indexes and revised application-search queries to reduce unnecessary scans."

Better when benchmark exists: "Reduced median application-search latency from X to Y in a documented benchmark by adding compound indexes and revising query stages."

Output

For each bullet provide:

• final wording;
• evidence source;
• ownership status;
• metric provenance (if any);
• likely interview follow-up question.

Guardrails

• No mandatory metric per bullet.
• No conservative guessing of percentages.
• No user/traffic/uptime/latency claims without evidence.
• No production language for development projects.
• No unsupported causal wording ("resulting in", "increased", "reduced") unless measured or otherwise established.