If you're a student or dev in India right now, you know the grind. I've spent months obsessing over my resume, hunting for keywords, and sending out countless applications for AI/ML roles. I did everything "right." The cold emails, the startup hunting, the works.
And what do you get in return? Either complete silence, or worse, a website asking you to PAY for an internship. ₹1,499 for a "1-Month AI Internship." As if the job market wasn't already stacked against us.
But after months of this, I realized the problem wasn't my resume. The problem was that I was applying to "ghosts”.
I started digging and realized that a massive chunk of job listings some estimate up to 90% in certain sectors are ghost listings. Companies post them just to collect our data, gauge salary expectations, or build a "talent pipeline" for a role that doesn't actually exist. It’s a waste of our time and a hit to our confidence.
I’m an AI student. Why am I not using my skills to fix this?
I decided to stop just "applying" and start "building."
I’m currently developing a Chrome extension designed to give job seekers a real-time "Credibility Score" for every listing they see on LinkedIn.
How it works (The Tech Behind the Logic):
I didn't want to build just another "AI wrapper." I’m building a system that aggregates five distinct signal categories to calculate a legitimacy score:
Company Legitimacy:
Cross-referencing domains and registration data.Recruiter Credibility:
Analyzing the profile behind the post.Posting Behavior:
Tracking how long a post stays up vs. actual hiring signals.On-Device ML:
I’m planning to use ONNX Runtime Web to run text classification locally in the browser to keep it fast and privacy-first.Why an extension and not a native app?
LinkedIn’s official APIs are locked behind enterprise gates. To help people now, we need to be where the jobs are. By using a Manifest V3 extension, I can provide an instant "Truth Layer" directly on the page without waiting for a formal partnership that may never come.The Goal:
I’ve previously built production AI systems that analyzed over 568k reviews and engineered RAG pipelines from scratch. Now, I’m applying that same "production-first" mindset to help students like me avoid the ghost-job trap.
I’m looking for collaborators!
This is still a work in progress. I’m currently mapping out the Vercel Edge functions and refining the ML text classifier.
If you're tired of ghost listings and want to help build this, connect with me on LinkedIn or drop a comment below.
Let’s stop applying into thin air :)
Top comments (1)
I really appreciate the intention behind this—a lot of students are genuinely struggling with signal‑to‑noise in the job market, and it’s good to see people trying to build solutions instead of just accepting the status quo.
One thought I’d offer, gently:
Ghost listings are a much more complex ecosystem than they appear on the surface.
Some are scams, some are “evergreen” roles, some are pipeline‑building, some are compliance artifacts, and some are simply abandoned postings. Because of that, the problem isn’t just technical—it’s also incentive‑driven, regulatory, and platform‑level.
Your extension idea is interesting, and the signal categories you listed are a good starting point. You might get even stronger results if you also explore:
I’m saying this with respect: the problem is solvable, but it may require a broader architecture than a classifier + heuristics. If you keep expanding the model of the ecosystem, your solution will get sharper and more durable.
Wishing you momentum on the build—it’s an important space, and your initiative is admirable.