How ResuBlue Works

1. What AI Powers ResuBlue?

• Model: Groq running Llama 3.3 70B Versatile
• Server-side only — the API key is never exposed to the browser. All AI calls go through ResuBlue's own servers.
• No BYOK — there is no “Bring Your Own Key” option. One provider, no user selection required.
• Together AI is configured as a server-side fallback only and is never shown to users.
• Groq data retention: transient inference — Groq does not train on your data. Groq Privacy Policy ↗

2. How We Prevent Hallucination (Guided Input Model)

• User provides facts; AI only structures what was provided. ResuBlue does not give the AI access to the internet, your LinkedIn, or any external data source.
• Every bullet-generation prompt includes the explicit instruction: “Use ONLY the facts provided. Do not add, invent, or infer any details.”
• Post-generation fact checker scans AI output for numbers and proper nouns that were not present in your original input.
• A facts-used panel shows a ✓ or ✗ for each input fact after bullet generation so you can verify what the AI used.
• A verification prompt appears before any AI content is inserted: “Does this accurately reflect your experience?”
• Every AI output in the app carries an AIGeneratedLabel badge so you always know which content was AI-assisted.

Important: These safeguards reduce errors but cannot eliminate them. Review all AI output before submitting to employers.

3. What the Proofreader Checks

The proofreader runs 45+ rule-based checks and 3 AI-powered checks. Here is what each category covers.

Formatting & Style

• Spelling — detects misspelled words throughout your resume.
• Tense consistency — flags mixed past and present tense within the same role.
• Active voice — identifies passive constructions that weaken bullet impact.
• Date format — catches inconsistent date styles (e.g. mixing “Jan 2024” and “2024-01”).
• Date format inconsistency — flags variation in how date ranges are presented across different roles.
• Bullet punctuation — warns when bullet endings are inconsistent (some with periods, some without).
• Filler phrases — removes weak language like “responsible for,” “helped with,” and “participated in.”
• Metrics present — checks whether bullets include quantifiable results (percentages, record counts, time saved).
• Summary length — warns if your professional summary is too short (<30 words), too long (>80 words), or starts with “I.”
• Led inflation — detects overuse of “led” when weaker verbs are more accurate.
• Keyword stuffing — flags unnaturally dense repetition of the same term, which ATS systems can penalize.
• Cover letter style mismatch — identifies tone or formatting inconsistencies between your resume and cover letter.
• Mobile formatting issues — highlights layout problems that make the resume hard to read on small screens.
• AI fingerprint detected — flags phrases that are statistically common in AI-generated text and may signal inauthenticity to recruiters (e.g. “spearheaded,” “leveraged,” “transformative”).

Salesforce-Specific

• Certification names — verifies that certification titles match current official Salesforce naming exactly.
• Salesforce capitalization — corrects improper casing of product names (e.g. “sales cloud” → “Sales Cloud”, “agentforce” → “Agentforce”).
• Deprecated terms — flags terminology retired by Salesforce, such as Process Builder (use Flow) and Workflow Rules (use Flow), and suggests current replacements.
• 2026 keywords — checks for high-value Salesforce skills currently in demand (Agentforce, Data Cloud, Atlas Reasoning, etc.).
• Certification expiry — warns when a listed certification may be approaching its maintenance deadline.
• Generic Salesforce usage — identifies vague claims like “used Salesforce” that fail to convey actual expertise.
• Org context — checks whether descriptions clarify the type of Salesforce org and its scale.
• Trailblazer profile link — suggests adding your Trailhead profile URL if it is missing from the contact section.
• DevOps maturity — flags missing mention of DevOps tools (Copado, Gearset, SFDX) for senior roles where that experience is expected.
• Spring '26 renames — catches product names updated in Spring 2026 (e.g. Connected Apps → External Client Apps).
• Cloud name outdated — identifies cloud product names that have since been renamed or consolidated.
• Cloud saturation — warns when a resume lists an unrealistically large number of Salesforce clouds without adequate detail.
• Architect inflation — flags use of “Architect” in a title or bullet not supported by corresponding certifications or responsibilities.
• Document AI requires Data Cloud — flags claims about Document AI without mention of Data Cloud, which is a prerequisite.
• Agent script open-source inflation — catches overclaiming of Agentforce contributions based on open-source templates.
• Headless 360 vague claim — flags imprecise descriptions of Headless or Connected Experiences without implementation context.
• Agentforce Grid beta claim — warns when the Agentforce Grid is cited as production experience when it remains in beta.
• Connected Apps deprecated pattern — identifies use of legacy Connected Apps authentication patterns replaced by newer OAuth flows.
• Agentforce Curated Memory pilot — warns when Curated Memory is described as a shipping feature rather than a limited pilot.
• everse / PISA research only — flags claims of production experience with technologies that are currently research-stage only.
• AI Associate retired — alerts that the Salesforce AI Associate certification was retired and is no longer a valid credential to list.
• Generalist red flag — identifies resumes that spread too broadly across unrelated Salesforce domains without a clear specialty.
• Data quality undersold — prompts users to highlight data quality and governance work, which is frequently underrepresented.
• Cert possibly expired — warns when a certification listed without a maintenance date may have lapsed.

ATS Compatibility

• Contact completeness — confirms your name, email, phone, and LinkedIn URL are all present and correctly formatted.
• ATS headings — verifies section headings match labels that ATS parsers recognize (e.g. “Experience” not “My Journey”).
• Emoji in content — warns that emojis in bullet points or headings can break ATS parsing.
• Special characters — flags decorative symbols and non-standard bullets that ATS software may misread.
• Unexplained gap — flags employment gaps longer than a threshold and suggests adding a brief explanation.
• Job hopping vs. consulting — distinguishes between instability and a legitimate consulting career pattern.
• Consulting career not formatted — flags consulting histories that should be formatted as a consulting block rather than separate employer entries.
• Overqualification risk — notes when experience level signals may cause automatic rejection by entry-level or mid-level ATS filters.
• Age signal detected — identifies details (early graduation years, obsolete technologies) that can inadvertently reveal age to screeners.
• Graduation year visible — warns when graduation year is present and may trigger age-based filtering by automated screeners.
• LinkedIn / resume mismatch — flags discrepancies between your resume dates or titles and what is publicly visible on LinkedIn.
• Bootcamp as employer — warns when a training bootcamp is listed in the experience section rather than the education section.
• Demo org as real experience — flags descriptions that appear to cite sandbox or demo org work as production client experience.
• Missing community section — suggests adding Trailblazer Community or user group involvement, which Salesforce recruiters actively look for.

Career Positioning & Authenticity

• AI authenticity — holistic AI-writing detection scoring your resume across phrase density, sentence variety, specificity depth, tonal consistency, and human voice signals.
• Low application ROI warning — identifies resume patterns that historically correlate with low interview-to-application ratios.

4. What AI Cannot Do

• Cannot verify your certifications or employment history.
• Cannot guarantee ATS compatibility — PDF export is image-based (html2canvas → canvas → PDF). ATS software cannot read it. Submit DOCX for ATS portals.
• Cannot confirm accuracy of metrics you provide — you are responsible for truthfulness.
• Cannot replace a professional career coach or recruiter.

5. Your Data and AI Processing

• Resume data is sent to Groq for AI features only — never for training.
• Pasted job descriptions are transient — never stored in the database.
• LinkedIn text (if pasted for consistency check) is transient — never stored. ResuBlue never fetches from LinkedIn directly.
• See our Privacy Policy for full data handling details.

6. ATS Export Honesty

• PDF export is image-based. ResuBlue renders the resume to a canvas via html2canvas, then converts that canvas to a PDF. The resulting file is a raster image — ATS software (Taleo, Workday, iCIMS, Greenhouse, and similar) cannot parse text from it.
• Use DOCX for ATS portals. The DOCX export produces structured, machine-readable text with proper heading hierarchy, skills tables, experience sections, education, and certifications.
• ResuBlue does not claim its PDFs are “ATS-safe” or “ATS-parseable.” When in doubt, submit DOCX.