🎯 CareerEngine: Open Source Autonomous Career Engine

📖 Setup & User Documentation: https://developerosindia.github.io/career-engine/
🌐 CareerOS Web App (Application Tracker): https://careeros.developeros.in/

🌐 The developerOS Career Suite

developerOS is a suite of free tools for engineers' careers, covering learning, tracking and applying:

CloudPrep: Teaches cloud certifications hands-on (AWS, GCP, Azure).
ResumeOS: Tailors your ATS resume with semantic AI matching.
CareerOS: Tracks applications, interviews, and offers in ₹ lakhs.
DevTrack: Analyses developer engineering activity and git metrics.
Insights: The engineering blog and industry salary benchmarks.
Toolkit: Browser-native utilities for everyday developer workflows.
CareerEngine (This Repository): The open-source automation layer: it applies to jobs through your own Chrome and runs on your machine. CareerOS is where you then track and compare what comes back.

ℹ️ Everything shares one developerOS sign-in, one design system and the same dark palette. CareerEngine is a standalone tool and does not sync with CareerOS yet.

⚠️ Important Notice on Platform Terms & Account Risk

Automating job applications can breach the terms of service of platforms like LinkedIn and Naukri, and risks account restriction. CareerEngine uses authentic Google Chrome CDP remote debugging with human-like delays to minimize detection, but we strongly advise:

Always run with --preview or --dry-run first to inspect forms before submitting.
Keep application volume conservative (5–15 jobs per session).
Do not run continuous unattended high-frequency loops.

🔒 Local Privacy Architecture & Gemini API

What stays strictly local: Your candidate master profile (profile_config.json), resume PDFs, API keys (.env), and SQLite database (jobs.db) remain entirely on your local machine and are protected by git pre-push security hooks.
What uses the cloud: Only the raw text prompts of individual job screening questionnaires are evaluated via the Google Gemini API keys that you provide in .env.

🚀 Key Features

CDP Remote Debugging Integration: Controls your authentic Google Chrome browser over port 9222, inheriting your active LinkedIn and Naukri sessions, cookies, and 2FA credentials.
Dynamic Form Intelligence:
- Automatically parses screening questionnaires across dropdowns (<select>), radio buttons, text fields, and file upload modals.
- Rotational Multi-LLM Engine: Round-robin rotation across multiple Google Gemini 3.x Flash-lite API keys (gemini-3.5-flash-lite, gemini-3.1-flash-lite) with DeepSeek fallback to prevent rate limits.
- Permanent QA Memory: Caches answered questions locally in profile_config.json to resolve recurring prompts instantly without redundant API calls.
- Form Validation Guardrails: Restricts numeric experience inputs to prevent form validation length errors (maxlength=20).
Contextual Multi-Role Resume Selector:
- Analyzes the job title and description in real-time and attaches the best matching resume PDF from resumes/ (_DE, _DS, _AI, _SWE, _MLOps).
Recruiter InMail Auto-Reply Engine:
- Inspects incoming LinkedIn messaging threads and drafts personalized responses highlighting candidate strengths, confirmed notice period (e.g. 30 days), and expected compensation.
8 Fully Supported Platforms:
- LinkedIn: Easy Apply automation, multi-step screening Q&A, and Recruiter InMail auto-responder.
- Naukri: Batch Quick Apply, keyword and location filtering, chatbot question solver.
- Wellfound (AngelList): Startup opportunities with dynamic personalized cover pitch notes.
- Instahyre: Curated tech openings with automated compensation and experience solving.
- Indeed: Global and India Easily Apply modal automation, resume uploading, and dry-run preview verification.
- Glassdoor: Easy Apply automation, salary expectations, and screening form completion.
- Foundit (Monster India): Quick Apply pipeline with instant dry-run audits and job card pagination.
- Hirist: Premium tech & startup opportunities in India with automated experience matching and one-click application support.
AI Cover Letter & Executive Memo Generator:
- Synthesizes personalized, high-converting cover letters across 3 tones (startup, enterprise, concise).
- Injects candidate metrics, solves hiring manager salutations, matches ATS keywords, and saves to cover_letters/ in Markdown and plain text.
Salary & Offer Negotiation Copilot:
- Benchmarks current vs offered compensation against empirical tech market distributions (india_inr_lpa, us_remote_usd_k, eu_remote_eur_k).
- Evaluates offers across P25, P50, P75, and P90 market bands.
- Generates 3 tactical counter-offer scenarios: standard counter, competing offer leverage, and total rewards pivot (sign-on bonus, equity/RSUs, remote flexibility).
- Supplies word-for-word email negotiation templates and live telephone talking points.
Stealth & Bot-Detection Safety Auditor:
- Pre-flight diagnostic engine checking Chrome CDP profile integrity, circadian pacing compliance (09:00–19:00), randomized jitter, session health, and daily velocity caps.
- Computes a comprehensive 0–100 Safety Score with risk classification (OPTIMAL_STEALTH, MODERATE_RISK, HIGH_RISK).
ATS Resume Keyword Tailor & Achievement Bullet Synthesizer:
- Compares any Job Description against your candidate profile facts and role tags.
- Detects high-priority keyword gaps and calculates baseline vs optimized ATS match scores.
- Synthesizes 3-4 quantified Google/Amazon XYZ-format achievement bullets and bespoke recruiter outreach pitch notes via multi-provider AI (Claude, OpenAI, Gemini, NVIDIA).
Application Pipeline Analytics & Multi-Format Exporter:
- Computes conversion rates, daily velocity, platform distribution, and top targeted companies.
- Exports tracking history to standardized CSV, structured JSON, and executive Markdown reports.
Relational Local History:
- Stored in SQLite database (jobs.db) with preview screenshots saved in screenshots/.

📁 Modular Directory Structure

career-engine/
├── docs/                             # GitHub Pages Documentation Website (developerosindia.github.io/career-engine)
│   ├── index.html                    # Responsive documentation & interactive playground
│   ├── styles.css                    # Glassmorphism dark mode aesthetic
│   ├── script.js                     # Tab switching, copy code, config generator
│   └── .nojekyll                     # Direct GitHub Pages asset serving
├── resumes/                          # Multi-Role Resumes Folder
│   ├── copy_your_resume_here.txt     # Placeholder & role tagging instructions
│   └── README.md                     # Documentation for _DE, _DS, _AI, _SWE tags
├── .env                              # Private API keys and local configuration (GITIGNORED)
├── .env.example                      # Public template for environment variables
├── .gitignore                        # Git exclusion rules (protects credentials and databases)
├── LICENSE                           # Open-source MIT License
├── config.yaml                       # Application settings, keywords, and search preferences
├── profile_config.json               # Candidate master profile facts & learned memory
├── main.py                           # Unified CLI entrypoint (apply, inmail, status, setup)
├── setup_agent.py                    # Interactive profile & secrets configuration wizard
├── agent.py                          # Backward-compatible wrapper for LinkedIn agent
├── requirements.txt                  # Python dependencies
│
├── core/                             # Core framework infrastructure
│   ├── __init__.py
│   ├── config_loader.py              # Loads .env & YAML with strict validation
│   ├── browser_engine.py             # Playwright CDP connection & stealth Chrome manager
│   ├── tracker.py                    # SQLite (jobs.db) & CSV (applied_jobs.csv) tracking
│   ├── models.py                     # Typed dataclasses (JobPosting, ApplicationRecord, InMailMessage)
│   ├── cover_letter.py               # AI Cover Letter & Executive Memo Generator
│   ├── salary_negotiator.py          # Salary & Offer Negotiation Copilot
│   ├── stealth_auditor.py            # Stealth & Anti-Bot Safety Pre-Flight Auditor
│   ├── ats_tailor.py                 # ATS resume keyword matching & XYZ bullet engine
│   ├── dedup_auditor.py              # Cross-platform SHA-256 deduplication
│   ├── recruiter_outreach.py         # Recruiter cold email & LinkedIn InMail synthesizer
│   ├── interview_prep.py             # System design & behavioral STAR drill generator
│   └── studio/                       # Local Web Studio GUI (port 8501)
│       ├── server.py                 # Python ThreadingHTTPServer REST API
│       └── static/                   # Pure CSS/JS dark-mode dashboard & Kanban
│
├── ai/                               # Multi-LLM reasoning engine
│   ├── __init__.py
│   ├── ai_solver.py                  # Rotational Gemini 3.x Flash-lite + NVIDIA DeepSeek + Claude + OpenAI
│   └── prompts.py                    # Structured system prompts (Form QA, InMail, Negotiation)
│
├── platforms/                        # Multi-platform job board architecture
│   ├── __init__.py
│   ├── base.py                       # BasePlatform abstract contract
│   ├── linkedin/                     # LinkedIn Easy Apply implementation
│   ├── naukri/                       # Naukri.com automated batch applicant & chatbot solver
│   ├── wellfound/                    # Connector for Wellfound / AngelList
│   ├── instahyre/                    # Connector for Instahyre curated tech openings
│   ├── indeed/                       # Connector for Indeed Apply
│   ├── glassdoor/                    # Connector for Glassdoor Easy Apply
│   ├── foundit/                      # Connector for Foundit (Monster India)
│   └── hirist/                       # Connector for Hirist India tech jobs
│
├── messaging/                        # Recruiter InMail auto-reply engine
│   ├── __init__.py
│   ├── inmail_scanner.py             # Scans /messaging/ for recruiter threads & InMails
│   ├── message_drafter.py            # Synthesizes personalized, professional replies
│   └── auto_responder.py             # Review & send coordinator (preview vs live send)
│
├── scripts/                          # Utility & launch scripts
│   └── launch_chrome.sh              # Launches Chrome on port 9222 with isolated profile
│
└── .agents/skills/career-automator/   # Agent Skill definition for Antigravity / Claude
    └── SKILL.md

⚙️ Installation & Setup

1. Clone & Install Dependencies

git clone https://github.com/developerOSindia/career-engine.git
cd career-engine

# Install required Python packages
pip install -r requirements.txt

2. Interactive Profile & Secrets Setup Agent

Run the interactive CareerEngine Setup Agent to automatically configure your profile, resume, search keywords, and API keys:

python3 main.py setup
# Or directly:
python3 setup_agent.py

The setup agent will walk you through setting up your:

Full name, contact details, and location
Resume PDF path
Target job keywords and experience years
Current and expected compensation (CTC)
Free Gemini API keys (from Google AI Studio)
Local configuration (.env, profile_config.json, config.yaml, jobs.db)

Note: All generated configuration files are strictly gitignored to keep your personal data local.

3. Launch Google Chrome in Remote Debugging Mode

Run the launcher script in a separate terminal:

./scripts/launch_chrome.sh

Note: If you are not already logged into LinkedIn in this Chrome window, log in once. Your session will remain safely cached in ~/.chrome-job-agent across restarts.

💻 CLI Usage Guide

1. Live Job Applications (`apply`)

Preview Applications in Dry-Run Mode (Safe Test):

Navigates to matching jobs, fills every form step, attaches resume, captures review screenshots, and pauses without clicking submit:

python3 main.py apply --platform linkedin --dry-run --limit 5

Auto-Submit Live Applications:

Applies and submits applications directly:

python3 main.py apply --platform linkedin --live --limit 10

Search Specific Roles & Locations:

python3 main.py apply --keyword "Databricks Engineer" --location "Bengaluru" --live --limit 5

Autonomous Batch Applications on Naukri (No URLs Required):

Automatically queries Naukri across keywords (Senior Data Engineer, Databricks Engineer, AWS Data Engineer, etc.) in Bengaluru/Remote, scans 20 job cards per page, skips already applied jobs, fills applications via Quick Apply, and answers chatbot screening questions:

# Preview batch of 20 Naukri jobs in dry-run mode:
python3 main.py apply --platform naukri --dry-run --limit 20

# Live auto-submit 20 jobs on Naukri:
python3 main.py apply --platform naukri --live --limit 20

# Filter by custom keyword & location:
python3 main.py apply --platform naukri --keyword "Databricks Engineer" --location "Bengaluru" --live --limit 20

Apply to a Specific Direct URL (Naukri or LinkedIn):

Apply directly to any specific job link (such as top-tier email alerts or Aurus early-jobs):

# Preview application on a direct Naukri job link:
python3 main.py apply --platform naukri --url "https://www.naukri.com/aurus/early-jobs?jobId=290626926058..." --dry-run

# Live auto-apply on a direct Naukri job link:
python3 main.py apply --platform naukri --url "https://www.naukri.com/aurus/early-jobs?jobId=290626926058..." --live

Autonomous Applications on Wellfound & Instahyre:

# Preview Wellfound startup applications with dynamic pitch notes:
python3 main.py apply --platform wellfound --dry-run --limit 5

# Live auto-submit applications on Wellfound:
python3 main.py apply --platform wellfound --live --limit 5

# Preview curated applications on Instahyre with automated screening Q&A:
python3 main.py apply --platform instahyre --dry-run --limit 5

Autonomous Applications on Indeed, Glassdoor, Foundit & Hirist:

# Preview Indeed Easily Apply jobs in dry-run mode:
python3 main.py apply --platform indeed --dry-run --limit 5

# Live auto-submit Indeed applications:
python3 main.py apply --platform indeed --live --limit 10

# Preview Glassdoor Easy Apply jobs:
python3 main.py apply --platform glassdoor --dry-run --limit 5

# Preview Foundit (Monster India) Quick Apply jobs:
python3 main.py apply --platform foundit --dry-run --limit 5

# Preview Hirist tech job openings:
python3 main.py apply --platform hirist --dry-run --limit 5

Apply to All 8 Supported Platforms in Round-Robin Batch:

python3 main.py apply --platform all --dry-run --limit 10

2. ATS Resume Keyword Tailoring (`tailor`)

Analyze any job description against your candidate profile facts, calculate baseline vs optimized ATS match scores, detect missing high-priority technical keywords, and synthesize 3-4 quantified Google/Amazon XYZ-format achievement bullets and personalized recruiter pitch notes:

# Tailor resume by pasting job description text:
python3 main.py tailor --title "Senior Backend Engineer" --company "Stripe" --jd "Looking for Python, FastAPI, PostgreSQL, Distributed Systems, Redis, and Kubernetes..."

# Tailor resume from a local JD text file:
python3 main.py tailor --title "Lead Data Engineer" --company "Snowflake" --file path/to/job_description.txt

# Tailor resume directly for an already tracked application in jobs.db:
python3 main.py tailor --job-id "ind_104928"

3. Application Pipeline Analytics & Export (`export` & `analytics`)

Export your full job application history from SQLite (jobs.db) to standard formats for spreadsheet analysis or team sharing:

# Export all applications to CSV:
python3 main.py export --format csv

# Export to structured JSON:
python3 main.py export --format json

# Export an executive Markdown pipeline report:
python3 main.py export --format markdown --output weekly_report.md

# View real-time pipeline conversion rates and platform breakdown:
python3 main.py analytics

4. Local Web Studio Dashboard GUI (`studio`)

Launch the zero-dependency developerOS Web Studio dashboard in your browser to inspect database metrics, preview form screenshots, manage candidate profile memory, and trigger CLI runs:

# Launch Studio on default port 8501:
python3 main.py studio

# Specify custom port or bind address:
python3 main.py studio --port 8599 --host 127.0.0.1

Dashboard: Real-time KPI counters (Applied, Dry Run Completed, Failed), status breakdown, and recent applications.
Applications Explorer: Searchable, filterable table of every job record with direct URLs, salary notes, timestamps, and 1-click Export to CSV.
Resume Tailor: Dedicated interactive ATS keyword optimizer, bullet generator, and recruiter pitch writer.
Run Orchestrator: Select platform (LinkedIn, Naukri, Wellfound, Instahyre, Indeed, Glassdoor, Foundit), mode (Live or Dry Run), and application limit with real-time terminal output streaming.
Screenshot Lightbox: Visual gallery of review screenshots captured during --preview and live runs.
Profile JSON Editor: Live viewer and editor for profile_config.json candidate facts and learned screening memory.
100% On-Device: Built natively on Python's standard library ThreadingHTTPServer with zero cloud telemetry or data leaks.

5. Autonomous Auto-Pilot Mode (`auto`)

Run CareerEngine on autonomous background auto-pilot with human circadian pacing, daily safety quota limits, multi-platform round-robin cycling across all 7 supported job boards, and automatic recruiter InMail answering:

# Launch Auto-Pilot in preview mode with safe defaults (180 min interval, 20 daily apps cap):
python3 main.py auto --interval 180 --daily-limit 20 --dry-run

# Launch Auto-Pilot in live auto-submit mode:
python3 main.py auto --interval 120 --daily-limit 15 --live

# Custom cycle interval and 24/7 unrestricted hours:
python3 main.py auto --interval 60 --session-limit 3 --all-hours --dry-run

Circadian Pacing: Only operates during business hours (09:00 - 19:00 local time) by default to avoid unnatural overnight bot activity.
Daily Safety Quotas: Enforces hard caps (e.g. 20 applications/day total) in SQLite jobs.db. Once reached, rests until the next day.
Round-Robin Multi-Platform: Cycles evenly across LinkedIn, Naukri, Wellfound, Instahyre, Indeed, Glassdoor, and Foundit.
InMail Monitoring Loop: Automatically inspects and answers unread recruiter inquiries after each batch.
Randomized Jitter: Varies sleep intervals by ±10% to eliminate static cron fingerprints.
GUI Controls: Can also be started, paused, and monitored visually from the Web Studio dashboard (http://localhost:8501).

6. Recruiter InMail & Message Auto-Reply (`inmail`)

Preview InMail Drafts (Dry-Run):

Scans your unread LinkedIn messages, analyzes recruiter inquiries, and displays proposed replies in the terminal without sending:

python3 main.py inmail --preview --limit 5

Live Auto-Send Replies:

Types and delivers AI-crafted responses directly into the message thread:

python3 main.py inmail --send --limit 5

Scan All Conversations (Including Read):

python3 main.py inmail --preview --all

7. Application Status Dashboard (`status`)

Inspect your complete application pipeline, status breakdown, and recent recruiter contacts:

python3 main.py status

Example Output:

============================================================
📊 CareerEngine - Application Status Dashboard
📁 Database: ./jobs.db
============================================================

📈 Applications Breakdown by Status:
  ✅ Applied             : 35
  🎯 Dry Run Completed   : 4
  ⏭️ Failed              : 7
  ⚠️ Needs Review        : 3
  ⏭️ Skipped             : 72

🕒 Last 10 Submitted Applications:
  • Full Stack Engineer with Mobile  @ Full Stack Engineer wi [Remote] | Hirer: Gayatri A
  • Software Engineer Intern (golang @ Software Engineer Inte [Remote]
  • Cloud Support Engineer @ Cloud Support Engineer [Remote]
  • Senior AI Compiler Engineer @ Senior AI Compiler Eng [Remote] | Hirer: Uday Bhaskar
  • Senior Statistical Programmer wi @ Senior Statistical Pro [Remote] | Hirer: Anchal Panchal
  • Linux Kernel Engineer @ Linux Kernel Engineer [Remote]
  • Software Engineer @ Software Engineer [Remote] | Hirer: Khursheed Ahmed
  • Remote Java Developer @ Remote Java Developer  [Remote]
  • (Senior) ServiceNow AI Engineer  @ (Senior) ServiceNow AI [Remote] | Hirer: Ravi prajapati
============================================================

8. Candidate Profile Scoring & ATS Audit (`score` / `audit`)

Audit and benchmark your candidate profile completeness, market readiness, and ATS fit before launching automated application campaigns:

# Run local deterministic ATS scoring and audit:
python3 main.py score

# Run scoring with AI executive critique from connected LLM (Claude, OpenAI, or Gemini):
python3 main.py score --llm

Evaluated Criteria (100 Points Total):

Contact & Online Presence (15 pts): Full name, verified email, telephone with country code, location anchor, and active LinkedIn / GitHub / portfolio links.
Work Experience & Compensation Fundamentals (25 pts): Current role, company, years of experience, metric-rich achievement descriptions, notice period calibration, and current vs. expected CTC in LPA.
Technical Skill Density & Recency (20 pts): Skill catalog breadth, core programming languages, relational/NoSQL databases, cloud infrastructure, and modern specializations.
Multi-Role Resume Tag Readiness (20 pts): Physical PDF resumes in resumes/ folder tagged with _DE, _DS, _AI, _SWE, and _MLOps for automated dynamic role switching.
Screening Memory & Automation Defaults (20 pts): Legal work authorization, visa sponsorship flags, joining availability, and cached custom_answers depth.

Produces an overall score, letter grade (A+, A, B+, B, C), multi-role fit % for 5 engineering tracks, and prioritized actionable recommendations to achieve 100%.

9. AI Cover Letter & Executive Memo Generator (`cover-letter`)

Generate high-converting, ATS-tailored cover letters that automatically pull your verified profile facts, match target role requirements, and export clean Markdown into cover_letters/:

# Generate a startup-tone cover letter:
python3 main.py cover-letter --company "Netflix" --title "Senior Distributed Systems Engineer" --tone startup

# Enterprise executive tone with specific hiring manager salutation:
python3 main.py cover-letter --company "Goldman Sachs" --title "Lead Cloud Architect" --hiring-manager "Sarah Jenkins" --tone enterprise

# Concise 3-paragraph executive memo:
python3 main.py cover-letter --company "Stripe" --title "Staff Software Engineer" --tone concise --save custom_letter.md

3 Calibrated Tones: startup (mission-driven, high agency), enterprise (structured impact, governance, scale), concise (tight executive pitch under 250 words).
Fact-Grounded: Uses your actual skills, experience, and key accomplishments from profile_config.json.
Zero Cloud Leak Fallback: If no AI API key is configured or offline, seamlessly falls back to a deterministic heuristic template generator.

10. Salary & Offer Negotiation Copilot (`negotiate`)

Data-driven compensation benchmarking and tactical counter-offer script generator. Evaluates candidate offers against real-world tech salary distributions across India, US, and European markets:

# Benchmark an offer for a Senior Backend Engineer in India (INR LPA):
python3 main.py negotiate --role "Senior Backend Engineer" --experience 6 --offered 34 --current 26 --market india_inr_lpa

# Benchmark a US Remote Staff Engineer offer (USD $k):
python3 main.py negotiate --role "Staff Software Engineer" --experience 9 --offered 195 --target 225 --market us_remote_usd_k

# Leverage a competing offer in tactical negotiation scripts:
python3 main.py negotiate --role "Lead Data Engineer" --experience 7 --offered 42 --competing 48 --market india_inr_lpa

Empirical Market Benchmarks: Built-in P25, P50, P75, and P90 compensation distributions across India (INR LPA), US Remote (USD $k), and EU Remote (EUR €k) for SWE, Data, Cloud, and Engineering Leadership.
Offer Health Rating: Classifies offers from BELOW_MARKET to TOP_PERCENTILE.
3 Tactical Action Plans:
1. Standard Counter-Offer: Assertive, grounded in target percentiles.
2. Competing Offer Leverage: High-urgency professional counter script.
3. Total Rewards Pivot: Structured trade-offs for sign-on bonus, equity/RSU vesting, and remote schedule flexibility when base salary is fixed.
Ready-to-Send Deliverables: Generates word-for-word email draft responses and bulleted telephone talking points.

11. Stealth & Bot-Detection Safety Auditor (`audit-stealth`)

Pre-flight diagnostic test checking your local automation environment to verify anti-bot compliance, humanized pacing, and Chrome CDP profile hygiene:

# Run local stealth and account safety audit:
python3 main.py audit-stealth

Audited Vectors:
- Chrome CDP Profile: Checks for authentic isolated user profile directory (~/.chrome-job-agent).
- Circadian Pacing: Verifies current time falls within human working hours (09:00–19:00 local time).
- Timing Jitter: Confirms randomized delay ranges (3.0s – 12.0s) to prevent static cron signature detection.
- Daily Velocity Quota: Audits jobs.db application count over the last 24 hours against safe daily caps (max 25/day).
- Platform Session Status: Checks for cached authentication cookies and active sessions across LinkedIn and Naukri.
Safety Score: Produces an overall 0–100 rating with clear risk tiers: OPTIMAL_STEALTH (85+), MODERATE_RISK (60–84), and HIGH_RISK (<60).

🧠 Semantic AI Matcher & Dynamic Pitch Generator

CareerEngine pairs candidate profiles with deep job description evaluation:

Semantic Scoring (0–100%): Uses Google Gemini to score role alignment against candidate experience before applying.
Resume Tag Alignment: Recommends the optimal resume tag (_DE, _DS, _AI, _SWE, _MLOps) and pairs it with matching resume PDFs.
Dynamic Startup Pitch Generator: Automatically synthesizes high-converting, personalized 2–3 sentence notes to founders and recruiters on platforms like Wellfound and LinkedIn InMail, emphasizing verified achievements and notice period without fluff.

📄 Multi-Role Intelligent Resume Selector

CareerEngine can automatically choose between multiple tailored resumes depending on the job you are applying to. Simply place your resume PDFs into the resumes/ folder using role tags:

Target Role	Filename Pattern	Examples
Data Engineering	`_DE.pdf`, `_data_engineer.pdf`	`resumes/Alex_DE_resume.pdf`
Data Science / ML	`_DS.pdf`, `_data_science.pdf`	`resumes/Alex_DS_resume.pdf`
AI / GenAI Platforms	`_AI.pdf`, `_genai.pdf`, `_llm.pdf`	`resumes/Alex_AI_resume.pdf`
Software Engineering	`_SWE.pdf`, `_software.pdf`, `_backend.pdf`	`resumes/Alex_SWE_resume.pdf`
MLOps / Cloud	`_MLOps.pdf`, `_devops.pdf`	`resumes/Alex_MLOps_resume.pdf`

CareerEngine analyzes the job title and description in real-time and attaches the best matching resume automatically.

👤 Candidate Profile & Customization

The candidate facts used to answer form questions and compose recruiter messages are defined locally in profile_config.json:

Metric / Fact	Default Example	Rules Enforced
Total Experience	Configurable (e.g. 5 yrs)	Used for experience verification
Current Role	Configurable	Verified from profile summary
Current CTC	Configurable (e.g. 30 LPA)	Fixed base + variable split
Expected CTC	Configurable (e.g. 40 LPA)	Enforced on compensation fields
Notice Period	Configurable (e.g. 30 days)	Answered across dropdowns & numeric inputs
Core Technical Skills	Configurable	Filtered against role requirements
Location Authorization	Configurable	Work authorization & relocation preferences

Dynamic Questionnaire Memory:

Whenever a new unique question is encountered, the AI generates the appropriate response based on your profile and records it into profile_config.json["custom_answers"]. You can edit or fine-tune any saved response directly in that file.

🛡️ Anti-Bot Safety & Rate Limiting

Realistic Delays:
- Dwell time on job pages (2.5s) before interacting.
- Delays between questionnaire steps (1.2s – 1.8s).
- Human keystroke jitter (30ms – 90ms per character).
- 3.5s cooldown between consecutive job applications.
Batch Processing: Designed to run in batches of 5 to 15 applications per session to keep account activity completely organic.
Safe Form Discard: Incomplete applications or unsupported multi-stage assessments are cleanly discarded using LinkedIn's Discard modal, preventing unwanted draft accumulations on your profile.

🌐 Multi-Platform Extension Guide

To implement a new job board (e.g. naukri, wellfound, indeed):

Navigate to platforms/<platform_name>/connector.py.
Inherit from BasePlatform and implement:
- authenticate_or_check_session(): Validates session cookies.
- search_jobs(): Extracts job listings and IDs.
- apply_to_job(): Automates form filling on that platform.
The platform is immediately runnable via python3 main.py apply --platform <name>.

⚖️ CareerEngine vs Legacy Bots vs Chrome Extensions

Feature / Architecture	CareerEngine (Open Source)	Legacy Python Bots (Selenium)	Browser Extensions
Execution Architecture	Authentic Chrome CDP (9222)	Headless / Selenium	In-page DOM Injection
Detection Evasion	Genuine Session & 2FA Cookies	Flagged / Triggers CAPTCHA	Detected by security scripts
Questionnaire AI	Rotational Gemini 3.x Flash-lite	Hardcoded regex lists	Single key / Paid cloud
Multi-Role Resume Selector	Contextual Auto-Routing (`_DE`, `_DS`, `_AI`)	Single static PDF	Static upload
InMail Auto-Reply	Autonomous Recruiter Negotiator	❌ Not Supported	❌ Not Supported
Data Privacy	100% Local (Git-safe, zero egress)	Plaintext credentials	Cloud-stored data
Cost	Free & Open Source (MIT)	Abandoned / Broken	$30–$80 / month

❓ Frequently Asked Questions (FAQ)

What is CareerEngine and how does it relate to CareerOS?

CareerOS (careeros.developeros.in) is the web platform where engineers track and compare job applications, interviews, and offers in ₹ lakhs. CareerEngine is the open-source automation companion: it runs locally on your machine, attaches to your Google Chrome browser via CDP port 9222, and assists with filling job applications across LinkedIn and Naukri. Note: CareerEngine is an independent open-source tool and does not sync automatically with CareerOS web yet.

Is there account risk when using automated tools?

Yes. Automated job applications can breach platform terms of service on LinkedIn and Naukri, and risk account restriction. CareerEngine controls your genuine Chrome browser with organic dwell times and typing jitter to minimize detection, but we strongly advise testing with --dry-run / --preview first, inspecting applications, and keeping application limits conservative (5–15 jobs per session).

Which AI models are supported for form filling?

CareerEngine features a multi-provider fallback pipeline:

Google Gemini (Gemini 3.5 Flash-lite / 3.1 Flash-lite via GEMINI_API_KEY_1..3)
Anthropic Claude (Claude 3.5 Haiku / Sonnet via ANTHROPIC_API_KEY or CLAUDE_API_KEY)
OpenAI (GPT-4o / GPT-4o-mini via OPENAI_API_KEY)
NVIDIA DeepSeek (DeepSeek V4 via NVIDIA_API_KEY)
Local Ollama (Offline zero-cost model via http://localhost:11434)

Learned answers are cached locally in profile_config.json["custom_answers"] to eliminate repetitive API latency and costs.

How does the Multi-Role Resume Selector choose a resume?

CareerEngine inspects the job title and requirements, matching them against tagged PDFs in resumes/ (e.g. _DE for Data Engineering, _DS for Data Science, _AI for GenAI Platforms, _SWE for Backend, and _MLOps for DevOps).

Is candidate data sent to the cloud?

Your candidate profile, resumes, API keys, and local SQLite database (jobs.db) stay strictly on your local machine. Only the raw questionnaire text prompts are sent to the LLM API keys that you configure in .env to evaluate answers.

Name		Name	Last commit message	Last commit date
.agents/skills/career-automator		.agents/skills/career-automator
ai		ai
core		core
docs		docs
messaging		messaging
platforms		platforms
resumes		resumes
scripts		scripts
tests		tests
.env.example		.env.example
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
config.example.yaml		config.example.yaml
launch_chrome.sh		launch_chrome.sh
main.py		main.py
profile_config.example.json		profile_config.example.json
requirements.txt		requirements.txt
setup_agent.py		setup_agent.py

Latest commit

History

Folders and files

Repository files navigation

🎯 CareerEngine: Open Source Autonomous Career Engine

🌐 The developerOS Career Suite

⚠️ Important Notice on Platform Terms & Account Risk

🔒 Local Privacy Architecture & Gemini API

🚀 Key Features

📁 Modular Directory Structure

⚙️ Installation & Setup

1. Clone & Install Dependencies

2. Interactive Profile & Secrets Setup Agent

3. Launch Google Chrome in Remote Debugging Mode

💻 CLI Usage Guide

1. Live Job Applications (apply)

Preview Applications in Dry-Run Mode (Safe Test):

Auto-Submit Live Applications:

Search Specific Roles & Locations:

Autonomous Batch Applications on Naukri (No URLs Required):

Apply to a Specific Direct URL (Naukri or LinkedIn):

Autonomous Applications on Wellfound & Instahyre:

Autonomous Applications on Indeed, Glassdoor, Foundit & Hirist:

Apply to All 8 Supported Platforms in Round-Robin Batch:

2. ATS Resume Keyword Tailoring (tailor)

3. Application Pipeline Analytics & Export (export & analytics)

4. Local Web Studio Dashboard GUI (studio)

5. Autonomous Auto-Pilot Mode (auto)

6. Recruiter InMail & Message Auto-Reply (inmail)

Preview InMail Drafts (Dry-Run):

Live Auto-Send Replies:

Scan All Conversations (Including Read):

7. Application Status Dashboard (status)

8. Candidate Profile Scoring & ATS Audit (score / audit)

9. AI Cover Letter & Executive Memo Generator (cover-letter)

10. Salary & Offer Negotiation Copilot (negotiate)

11. Stealth & Bot-Detection Safety Auditor (audit-stealth)