CareerAI - AI-Powered Career Placement Assistant

AI-powered career tool helping students ace placements with instant resume scoring, mock interviews, skill gap analysis, and personalized job matching - all free!

Comment 1

Inspiration

During my third year of college, I watched my seniors struggle during placement season. One friend had an amazing profile but got rejected from 15 companies without a single interview call. When I looked at his resume, I realized it was poorly formatted and missing key technical keywords that ATS systems look for.

Another senior aced technical rounds but failed behavioral interviews because he'd never practiced answering "Tell me about yourself" or "Why this company?" questions. Career coaching costs ₹20,000+ which most students in tier-2 colleges can't afford.

That's when it hit me - what if AI could democratize career prep for everyone? Not just students from IITs or those who can afford expensive coaching, but every student in Bharat trying to land their first job.

What it does

CareerAI is like having a career coach in your pocket. Here's what it offers:

Resume Checker - Upload your resume PDF and get an instant ATS compatibility score out of 100. The AI tells you exactly what's missing - whether it's specific keywords, better formatting, or clearer achievement statements. I tested it with my own resume and improved my score from 62 to 89 in one iteration!

Mock Interview Practice - Select any job role (software engineer, data analyst, product manager, etc.) and the AI generates realistic interview questions. Type your answer, hit submit, and get detailed feedback on what you did well and where you can improve. It's like doing mock interviews with a senior, but available 24/7.

Skill Gap Analysis - Ever looked at a job posting and wondered "Am I even qualified for this?" This feature extracts skills from your resume, compares them with what companies actually want, and shows you the gap. Better yet, it gives you a personalized learning roadmap with free YouTube courses and documentation links.

Job Recommendations - Instead of scrolling through thousands of Naukri listings, the AI matches jobs to your profile and ranks them by how good the fit is. It even tells you WHY each job matches and what you need to learn to be competitive.

How I built it

I started by sketching the UI on paper at 10 PM yesterday (yeah, classic hackathon move). By midnight, I had the React frontend skeleton up with Tailwind CSS for styling.

The trickiest part was PDF parsing - resumes come in all formats and extracting clean text was a nightmare. After trying 3 different libraries, I settled on pdf.js which handled most edge cases. Still not perfect with tables though!

For the AI backend, I used Groq's Mixtral API because it's fast and free (shoutout to Groq for making AI accessible). I spent hours crafting the right prompts - turns out getting consistent JSON responses from LLMs requires very specific instructions. The resume analysis prompt alone went through 12 iterations before I got reliable results.

I hooked everything up with Express on the backend, MongoDB for storing user data, and deployed the frontend on Vercel around 2 AM. Testing with my friends' actual resumes revealed bugs I never expected (special characters breaking the parser, anyone?).

The UI polishing happened in the last 4 hours - adding those gradient effects, smooth animations, and making sure it works on mobile. My phone became my best testing device at 3 AM.

Challenges I faced

PDF Parsing Hell - Different resume templates have wildly different structures. Some use tables, some use columns, some are basically images with text overlays. Getting clean, structured text out of these was brutal. I eventually added preprocessing to clean up extracted text, but it's still not 100% accurate.

Prompt Engineering - Making the AI give consistent, useful feedback took way longer than coding. Early versions would either give generic advice like "make your resume better" or hallucinate skills you never mentioned. I added strict output format requirements and examples to the prompts, which helped a lot.

Performance - First version took 15+ seconds to analyze a resume because I was calling the AI API multiple times sequentially. Optimized it to batch requests and cache common responses. Now it's under 5 seconds, which is acceptable.

Time Management - Classic hackathon problem. I wanted to add voice-based interviews and a Chrome extension, but had to cut scope to ship something functional. Prioritizing features when you're running on 3 hours of sleep is tough.

UI/UX Balance - Making it look professional without spending 10 hours on design. I kept iterating between "functional but ugly" and "pretty but broken" before finding middle ground around 3 AM.

What I learned

This was my first time doing serious prompt engineering with LLMs. I learned that AI is incredibly powerful but also surprisingly dumb if you don't guide it properly. The difference between a good prompt and great prompt is massive.

Also learned a ton about PDF handling, which is more complex than it seems. And deployment - Vercel makes it stupidly easy to ship React apps, though I did hit some build errors around environment variables.

On the soft skills side, I learned to cut scope aggressively. I had ideas for 10 features but shipped 4 really solid ones instead. Better to have 4 working features than 10 half-baked ones.

What's next for CareerAI

Immediate plans (if I win or even if I don't):

  • Add voice-based mock interviews using speech recognition
  • Support for Hindi and other regional languages (this is Bharat, after all)
  • Resume templates that are ATS-friendly by design
  • Mobile app because not everyone has laptops

Bigger vision:

  • Company-specific interview prep (Amazon Leadership Principles, Google's STAR format, etc.)
  • Integration with LinkedIn to auto-import profiles
  • Community features where students can share interview experiences
  • Partnership with colleges to offer this as a placement prep tool

Dream scenario: If this actually helps students get jobs, I want to make it sustainable. Maybe a freemium model where basic features stay free forever, but advanced stuff like 1-on-1 mentorship or detailed analytics become paid. But the core mission stays the same - democratize career prep for every student in India.

Right now though, I need to submit this before the 4 AM deadline and then sleep for 12 hours straight 😅

Built with ❤️ and way too much coffee during AI for Bharat 2026 Hackathon.

Built With

Share this project:

Updates

Submission history