Learn why we built this tool, how the AI works, and our promise to Indian job seekers.
Every single day, millions of Indians receive fake job offers on WhatsApp, email, Naukri, and LinkedIn. Scammers impersonate well-known companies like Infosys, TCS, Wipro, and even government ministries. They promise high-paying roles and then ask victims to pay a "registration fee," "security deposit," or "training cost" — often ranging from ₹500 to ₹50,000.
These scams target India's most vulnerable job seekers — freshers, people from small towns, those recently laid off, and anyone desperate for income. According to government cybercrime data, job fraud is one of the fastest-growing categories of online scams in India, with thousands of complaints filed every month.
Fake Job Detector India was built to give every job seeker — regardless of technical knowledge — a free, instant, and reliable way to verify any job post before responding, paying, or sharing personal information. We wanted a tool trained specifically on Indian job data, not just generic global datasets.
At the core of Fake Job Detector India is a Random Forest machine learning classifier. Random Forest is an ensemble method that builds hundreds of decision trees during training and combines their predictions for higher accuracy and robustness than any single decision tree.
Job post text is first converted into numerical features using TF-IDF (Term Frequency–Inverse Document Frequency) vectorization. TF-IDF identifies which words and phrases are especially significant in distinguishing fake jobs from real ones — for example, terms like "registration fee," "weekly payment," "WhatsApp only," and "no experience needed" score very high in fake job posts.
On top of the ML model, we apply a layer of India-specific rule-based detection that catches local scam patterns the global training data might miss: unrealistically high salary figures in rupees, mentions of specific fake IT company names, requests for Aadhaar or PAN upfront, and WhatsApp-only contact patterns.
The final result combines the ML probability score with the rule-based flags into a single Risk Score (0–100) and a verdict: Safe, Suspicious, or High Risk.
Our model was trained on a combined dataset of 96,674 labeled job postings:
The model achieves the following metrics on the held-out test set:
A note on the F1 score: the fake-job class is inherently imbalanced (real jobs outnumber fake ones). An F1 of 0.7789 on the fake class means our model correctly identifies most fake jobs while keeping false positives low — a practical balance for real-world use.
Fake Job Detector India was built by Amit Mastud, an independent developer passionate about using machine learning to solve real problems that affect everyday Indians. The project started as a personal initiative after observing how many people in his network had received — and sometimes fallen victim to — fake job scams on WhatsApp.
The project is open source. All code, including the training pipeline and Chrome extension, is available on GitHub. Contributions, suggestions, and bug reports are welcome.
If you've found a fake job post — whether our tool flagged it or you noticed it yourself — please report it to:
Reporting fake jobs helps authorities track fraudsters and protects other job seekers. It takes less than 5 minutes.