Job Post Red Flag Analyzer Job Posting Red Flag Analyzer
An AI-powered tool that analyzes job postings for scam red flags. Built on PartyRock/Amazon Bedrock, with a companion landing page on AWS Amplify.
Helping Fresh Graduates Spot Warning
Signs Before They Apply
The problem I experienced while job hunting
I've been applying for jobs as a fresher in India for the past year.
Every day, I spend a significant amount of time going through job postings across different platforms. Finding an opening is only the first step. The harder question is:
Can I actually trust this job posting?
I recently experienced this problem myself.
I applied for an internship posted by a company through an official national internship portal. Around 15 minutes after applying, I received an email saying that I had been shortlisted.
At first, I thought it might simply be a fast screening process.
Later, I searched for the company on LinkedIn and found someone describing their own experience with the same company. According to their post, candidates were asked to work without pay for a period of time, complete assignments, and submit the work by a deadline. After the work was completed, the company stopped responding to calls and emails.
That made me realize something important:
A job posting appearing on a trusted platform does not automatically mean that every aspect of the opportunity is trustworthy or fair.
For a fresher who is already dealing with intense competition for a limited number of opportunities, knowing what to look for can be difficult.
So I decided to build something around that problem.
What I built
I built Job Posting Red Flag Analyzer, an AI-powered tool designed to help fresh graduates identify warning signs in job postings before they invest more time or share sensitive information.
The user pastes the text of a job posting and can optionally provide the company name.
The application returns:
- A Trust Score from 0–10
- A Safe, Caution, or High Risk classification
- The specific red flags detected
- A plain-English explanation of each flag
- An overall verdict
- Suggested next steps for verification
The goal is not to tell someone:
"This company is a scam."
Instead, the goal is:
"Here are the warning signs in this particular posting. Here's why they matter, and here's what you should verify before proceeding."
That distinction became an important part of how I designed the application.
The six categories I check
I structured the analyzer around six categories of warning signs.
1. Money red flags
The application looks for:
- Upfront fees
- Registration or processing charges
- Deposits
- Mandatory purchases
- Payment requests
- Unrealistic salary claims for fresher roles
2. Company legitimacy
It checks for signals such as:
- Generic email domains
- Missing or vague company information
- No verifiable web presence
- Potential impersonation of a known company
- Claims that cannot be connected to an identifiable organization
These are treated as signals for further verification, not automatic proof that a company is fraudulent.
3. Urgency and pressure
The analyzer looks for:
- Artificial deadlines
- "Apply immediately" pressure
- Instant shortlisting without meaningful screening
- Pressure to move conversations to WhatsApp or Telegram
- Other language designed to make candidates act before checking the opportunity
4. Job description quality
It checks for:
- Vague responsibilities
- Buzzword-heavy descriptions
- Mismatch between the job title and actual duties
- Missing interview-process information
- Unrealistic expectations without meaningful role details
5. Personal information overreach
The application looks for requests for sensitive information such as:
- Aadhaar details
- PAN details
- Bank information
- Payment screenshots
- Irrelevant documents
The timing of such requests matters. Asking for sensitive information before a legitimate hiring stage deserves additional scrutiny.
6. Unpaid work and ghosting risk
This category came directly from the experience that motivated the project.
The analyzer looks for:
- Unpaid trial periods
- Performance-based payment before a formal offer
- Real deliverables disguised as an "assignment"
- Work being requested without clear employment terms
- Missing offer letters
- Patterns where candidates may complete substantial work without a clear commitment from the employer
How I built the AI workflow
I didn't want one large AI prompt to simply read a job description and produce a subjective "safe" or "unsafe" answer.
Instead, I structured the application as a four-stage pipeline.
Stage 1 — Signal Extraction
The first stage reads the raw job posting and extracts structured signals.
For example, it identifies whether the posting contains:
- Fee requests
- Sensitive-information requests
- Urgency language
- Contact methods
- Legitimacy signals
- Unpaid work requirements
- Other relevant indicators
This gives the rest of the workflow structured information to work with.
Stage 2 — Category Scoring
The extracted signals are evaluated against the six red-flag categories.
I used explicit rules to reduce false positives.
For example, a short job description should not automatically be considered suspicious. A small company having a limited online presence should not automatically be treated as fraudulent either.
The goal was to distinguish between:
"This deserves investigation"
and
"This proves the employer is fraudulent."
Those are very different claims.
Stage 3 — Trust Score Calculation
The application combines the category results into a Trust Score and risk classification.
I also introduced deterministic override rules for particularly significant combinations of signals.
For example, multiple independent risk categories can force a High Risk classification, while an unverified impersonation of a real company receives additional scrutiny.
This helped make the output more consistent than simply asking the AI to freely generate a score.
Stage 4 — Verdict and Next Steps
Finally, the results are converted into language that a job seeker can understand.
Instead of returning technical AI terminology, the application explains:
What was detected → Why it matters → What the candidate should check next.
The suggested actions can include checking the company's official website, verifying the opportunity through an official employer channel, checking relevant company information, or reporting suspected cybercrime through the appropriate government channel.
Where AWS fits
I built the core application using PartyRock, using Amazon Bedrock-powered generative AI capabilities.
PartyRock allowed me to build and iterate on the AI workflow without having to build a complete backend from scratch.
I also used Whiskers, PartyRock's AI building assistant, to help construct and iterate on the application.
The final workflow is intentionally more structured than a single AI prompt:
Job Posting → Signal Extraction → Category Analysis → Trust Score → Explanation & Next Steps
Alongside the core PartyRock application, I created a companion landing page using AWS Amplify Hosting.
Technology used
- PartyRock / Amazon Bedrock — AI-powered analysis
- Whiskers — AI-assisted application building and iteration
- AWS Amplify Hosting — companion landing page
This project was especially meaningful to me because I did not have extensive hands-on AWS experience before starting it.
I learned by building an actual application around a problem I personally experience rather than learning the services in isolation.
Testing the application
I didn't want to stop after getting convincing AI responses.
I tested the application against different types of job postings to see whether its behavior changed appropriately.
One test involved an internship offer I personally received from "a company I applied to", where the offer contained a stipend condition tied to training or project clearance. The analyzer classified it as High Risk with a 2/10 score.
I also tested an official TCS recruitment page, which received a 7/10 Safe classification.
Another useful test involved an official Accenture job posting from its careers website. The analyzer scored the official posting 10/10 Safe.
I then tested a LinkedIn repost referring to the same opportunity but containing shortened redirect links and unrelated spam content. That version received a 2/10 High Risk result.
This was important because the application was not simply learning:
"Accenture = safe."
It was responding to the signals contained in the particular version of the posting being analyzed.
Testing exposed problems in my own AI logic
One of the most valuable parts of building this project was discovering where my first versions were wrong.
An early version penalized legitimate companies simply because their job descriptions were short.
That was a problem.
A short job description is not automatically a scam.
I changed the logic and tested it again.
Later, while trying to avoid false positives against smaller businesses, I over-corrected and weakened the application's ability to identify potential impersonation of real companies.
I corrected that behavior and re-ran my test cases.
This changed the project from simply:
"I built an AI that gives a risk score."
into:
"I built an AI workflow, tested where it failed, changed the logic, and re-tested the changes."
That was one of my biggest learning experiences from this project.
What the application cannot do
There is an important limitation I want to make clear.
Job Posting Red Flag Analyzer analyzes the posting text provided by the user.
It does not automatically:
- Verify a company's registration in real time
- Check live company reviews
- Confirm that a recruiter actually works for the company
- Guarantee that an employer is legitimate
- Guarantee that an employer will treat a candidate fairly
- Determine that a company is fraudulent based only on the posting
The application is therefore a screening and awareness tool, not a replacement for independent verification.
It flags patterns worth investigating.
It does not claim certainty where the available information does not support it.
This was an important design decision because an AI-generated "Safe" result should never be treated as proof that a job is genuine.
Why I built it
When you've been applying for jobs for months, every opportunity can feel important.
Messages such as:
"You have been shortlisted."
"Complete this assignment immediately."
"Pay a small registration fee."
or
"Send your documents now."
can create pressure to act before thinking about whether the opportunity deserves scrutiny.
I wanted to build something that gives fresh graduates a moment to pause.
Not:
"Don't apply."
But:
"Here are the things you should check before you continue."
What I learned
This was my first serious hands-on experience building with AWS generative AI tooling.
I learned that building an AI application is not just about getting a model to produce an impressive response.
The harder questions are:
- What information should the model extract?
- Which signals actually matter?
- How should different signals interact?
- How do I reduce false positives?
- How do I avoid overclaiming what the model knows?
- How do I test whether a change actually improves the application?
PartyRock gave me a practical way to experiment with these questions while learning how generative AI workflows can be structured.
The project also taught me that responsible AI output is part of product design, especially when the output can influence someone's decision about a real job opportunity.
What's next
The current version is a working prototype, but I want to take the idea further.
A future production version could incorporate additional verification layers such as:
- Official company website verification
- Employer-domain verification
- Company registration data where appropriate
- Cross-checking official career pages
- Detection of duplicated or suspicious job descriptions
- Community reports and candidate experiences
- Better source-level verification
- Historical signals associated with previously reported postings
The long-term goal is not to replace a job seeker's judgment.
It is to give job seekers better information before they make that judgment themselves.
Try the project
Job Posting Red Flag Analyzer:
https://partyrock.aws/u/Keerthimorugu/ROIt2dbd7/Job-Posting-Red-Flag-Analyzer
Final thought
I started this project because I was tired of asking myself the same question every time I found a promising job:
"Is this actually worth applying to?"
I don't think an AI tool can answer that question with certainty.
But it can help identify warning signs that a fresher might otherwise miss.
And sometimes, giving someone a few extra signals before they share personal information, spend their time on unpaid work, or trust an opportunity too quickly can make a meaningful difference.
That is what I wanted to build with Job Posting Red Flag Analyzer.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article