JobLens

AI-powered job matching for Switzerland. Upload your LinkedIn profile, and JobLens scores hundreds of real job postings against your skills using the LLM of your choice.

Everything runs in your browser. Your API keys and profile data never touch a server -- you bring your own key (BYOK), and all processing happens client-side.

Features

• LinkedIn PDF parsing -- drag-and-drop your LinkedIn export, and an LLM extracts a structured profile (skills, experience, languages, education)
• jobs.ch integration -- searches Switzerland's largest job board via its semantic search API and scrapes full job descriptions
• LLM-powered scoring -- each job is scored 0-100 based on skill match (40%), experience level (25%), domain fit (20%), and language fit (15%)
• Multi-provider support -- works with OpenAI (GPT-4o), Anthropic (Claude Sonnet), or Google Gemini
• Filtering & sorting -- filter by minimum score, sort by match score, date, distance, or company
• Distance calculation -- shows how far each job is from your city (Haversine formula, 24 Swiss cities)
• Privacy-first -- no backend, no database, no tracking. API keys are stored only in your browser's localStorage

Getting Started

Prerequisites

• Node.js 18+
• An API key from OpenAI, Anthropic, or Google AI Studio

Run locally

# 1. Start the CORS proxy
cd worker
npm install
npm run dev          # runs on http://localhost:8787

# 2. Start the webapp (in a second terminal)
cd webapp
npm install
npm run dev          # runs on http://localhost:5173

Open http://localhost:5173 and follow the 4-step wizard.

Deploy

• Webapp: deploy the webapp/ directory to Vercel, Netlify, or GitHub Pages (npm run build produces a static dist/ folder)
• Proxy: deploy the worker to Cloudflare (cd worker && npm run deploy), then set VITE_PROXY_URL to your worker URL before building the webapp

How It Works

1. Settings -- pick your LLM provider and enter your API key
2. Profile -- upload your LinkedIn PDF; text is extracted client-side with pdf.js, then an LLM structures it into JSON
3. Search -- enter keywords (pre-filled from your profile title); JobLens fetches up to 2000 jobs from jobs.ch, enriches the top 50 with full descriptions, and batch-scores them with your LLM
4. Results -- browse scored jobs with color-coded match scores, matching/missing skills, and direct links to job postings

Architecture

                        +-----------------------+
                        |    User's Browser     |
                        |                       |
                        |  React SPA (Vite)     |
                        |  - PDF parsing        |
                        |  - LLM calls          |
                        |  - Job scoring         |
                        |  - Zustand state      |
                        +----------+------------+
                                   |
                          All requests via
                          /proxy/<host>/...
                                   |
                        +----------v------------+
                        |  Cloudflare Worker    |
                        |  (CORS Proxy)         |
                        |                       |
                        |  - Stateless          |
                        |  - No storage/logging |
                        |  - Host whitelist     |
                        +----+------+-------+---+
                             |      |       |
               +-------------+  +---+---+  +------------+
               |                |       |               |
        +------v------+  +-----v-----+ |  +------------v---+
        |  LLM APIs   |  | jobs.ch   | |  | jobs.ch        |
        |             |  | Search API| |  | Detail Pages   |
        | - OpenAI    |  +-----------+ |  | (JSON-LD)      |
        | - Anthropic |                |  +----------------+
        | - Gemini    |                |
        +-------------+                |
                                       |
                              +--------v--------+
                              | review-api      |
                              | .jobs.ch        |
                              +-----------------+

Webapp (webapp/)

Directory	Purpose
src/components/	4 step components (Settings, ProfileUpload, JobSearch, JobResults) + JobCard
src/services/	Core logic: llm-client (multi-provider LLM wrapper), job-scraper (jobs.ch API + HTML scraping), job-scorer (batch scoring), pdf-parser (pdf.js text extraction), geo (distance calc)
src/prompts/	LLM prompt templates for profile extraction and job scoring
src/store/	Zustand store with localStorage persistence (API key, profile, search params)
src/types/	TypeScript interfaces for Job, ScoredJob, UserProfile, etc.

CORS Proxy (worker/)

A ~70-line Cloudflare Worker that transparently forwards requests to whitelisted hosts. It exists solely to bypass browser CORS restrictions. It does not store, log, or transform any data. Allowed hosts:

• api.openai.com, api.anthropic.com, generativelanguage.googleapis.com (LLM providers)
• job-search-api.jobs.ch, www.jobs.ch, review-api.jobs.ch (job data)

Data Flow

1. PDF upload -- pdf-parser.ts extracts raw text using pdf.js (entirely client-side)
2. Profile extraction -- llm-client.ts sends text to chosen LLM with a structured extraction prompt; returns UserProfile JSON
3. Job search -- job-scraper.ts calls jobs.ch semantic search API (paginated, up to 20 pages of 100 jobs). For the top 50 results, it fetches detail pages and extracts descriptions from JSON-LD markup
4. Distance enrichment -- geo.ts calculates Haversine distance from the user's city to each job's coordinates
5. Scoring -- job-scorer.ts batches 20 jobs per LLM call (3 concurrent batches), scoring each job 0-100 with matching/missing skills
6. Display -- JobResults.tsx renders sorted/filtered results with JobCard components

Tech Stack

Layer	Technology
Frontend	React 19, TypeScript, Vite
Styling	Tailwind CSS
State	Zustand (with localStorage persistence)
PDF parsing	pdfjs-dist (client-side)
CORS proxy	Cloudflare Workers
LLM providers	OpenAI, Anthropic, Google Gemini
Job data	jobs.ch (search API + HTML scraping)
Testing	Vitest

Development

npm run dev          # start dev server (webapp or worker)
npm run build        # type-check + production build (webapp)
npm run test         # run tests (webapp)
npm run test:watch   # run tests in watch mode (webapp)
npm run lint         # ESLint (webapp)
npm run deploy       # deploy to Cloudflare (worker)

License

MIT