🎯 AI ATS Evaluator & Career Suite

An evidence-based ATS evaluator that scores resumes on weighted requirement coverage —
not arbitrary keyword-density deductions like most resume checkers.

🚀 Try it Live

---

What makes this different

Most ATS checkers count keywords and dock points blindly. This one runs a real weighted rubric: the top 3-4 requirements near the top of a job description carry 60% of the fit score, secondary/nice-to-have skills carry the remaining 40% — the same way a recruiter actually screens, not a flat keyword match.

It also catches a specific, common resume-writing failure most tools miss entirely: keyword stuffing without evidence. If a skill only appears in a bare "Skills" list with no supporting work-experience bullet, the evaluator downgrades its match weight to 0.5 and flags it as unverified — because a recruiter would catch that gap too.

Features

Module	What it does
📊 Recruiter Audit	Weighted fit score (0-100), critical/warning red flags, 3-5 non-negotiable knockout checks, missing core vs. bonus skills
🔍 Anti-Stuffing Detection	Downgrades and flags skills with no bullet-point evidence, instead of trusting the skills list at face value
💥 Impact Analysis	Flags banned filler words and weak bullets, rewrites them with the specific flaw explained
📝 Live Editor	Edit the extracted resume text directly and re-run evaluation
🎙️ Interview Prep	5 targeted questions (3 technical gaps + 2 behavioral) with ideal talking points, generated from your actual gaps
⚡ Outreach Generator	A cold email (under 100 words) and a tailored 4-paragraph cover letter — with an explicit banned-word list ("delve," "seamlessly," "spearheaded," "results-driven," "thrilled") to avoid sounding AI-written
📜 Application History	Persistent Supabase storage — search and revisit past evaluations
📄 Direct File Upload	Accepts PDF and Word documents directly — PDFs are sent natively to Gemini, not just text-extracted

Tech Stack

• Framework: Next.js (App Router), TypeScript
• AI Engine: Google Gemini API (gemini-2.5-flash), structured JSON schema outputs, temperature=0 for evaluation determinism
• File Parsing: Native PDF ingestion via Gemini, mammoth for .docx fallback
• Database: Supabase (Postgres)
• Styling: Tailwind CSS
• Deployment: Vercel

Known TODO

• Model version: currently pinned to gemini-2.5-flash. My other active projects run on the newer gemini-3.x line — this needs a bump and a re-test before it's fully current.

Getting Started

1. Clone and install

git clone https://github.com/Ananth-Sai/ATS-Evaluator.git
cd ATS-Evaluator
npm install

2. Set up Supabase

Run this in your Supabase SQL editor:

create table applications (
  id uuid default gen_random_uuid() primary key,
  created_at timestamp with time zone default timezone('utc'::text, now()) not null,
  job_title text,
  company_name text,
  match_score numeric,
  analysis_json jsonb
);

alter table applications enable row level security;

create policy "Allow public access to applications"
on applications for all
using (true)
with check (true);

3. Environment variables

Create .env.local:

GEMINI_API_KEY=your_gemini_api_key_here
NEXT_PUBLIC_SUPABASE_URL=your_supabase_project_url
NEXT_PUBLIC_SUPABASE_ANON_KEY=your_supabase_anon_key

4. Run it

npm run dev

Open http://localhost:3000.

Deployment

Deployed on Vercel — push to GitHub, import the repo, and add the same three environment variables in the Vercel project settings.

License

MIT