ATS Screener

Most free ATS checkers give you one invented number and lock the rest behind a paywall. ATS Screener does the honest, harder thing: it returns six separate scores from six real enterprise platforms, each tuned to how that vendor actually parses, filters, and ranks a resume.

Resumes (PDF or DOCX) are parsed fully client-side in a Web Worker with pdfjs-dist and mammoth, so the file itself never leaves the browser, only the extracted text is sent on. A custom NLP layer (tokenizer, TF-IDF, synonyms, and a multi-industry skills taxonomy) pulls sections, skills, and dates out of raw text.

A deterministic rule engine then scores against all six platform profiles, weighting formatting, keyword match, sections, experience, education, and quantification, plus per-platform quirks, and switching between exact, fuzzy, and semantic keyword matching to mirror how each real system behaves. On top of it sits a cross-provider LLM chain, Gemma 3 27B on Google falling back to Llama 3.3 70B on Groq, with the rule engine as a final fallback so the app keeps working even when every model provider is down.

Auth and storage adapt to where it runs: Firebase and Firestore for the hosted build, anonymous localStorage for self-hosters, or on-prem LDAP and Active Directory. A live, Firestore-backed counter on the landing page has tracked more than 2,000 people through it so far, and the full behavior is written up in its own documentation site.