How I Built AI Skill Matcher – A Career Path Platform Powered by Groq AI

Introduction

One question I kept hearing from people around me was: "I know I'm good at things. But how do I make money from them?" Generic career advice doesn't help here.
"Learn coding." "Start freelancing." "Build a personal brand." These answers are too broad.
They don't match the individual. So I decided to build something that actually listens — and gives a real answer. That's how AI Skill Matcher started.

The Problem

Most people don't lack skills.
They lack direction. The gap is between:

what someone is good at
and what they can actually do with it to earn income

Existing tools give generic results.
They don't understand your specific situation, time availability, or income goals. There was no tool that took your actual answers and gave you a personalized, actionable path.

The Idea

I wanted to build a platform where a user could:

answer a short quiz about their skills and interests
get three realistic income paths generated by AI
see a full step-by-step roadmap for each path
track progress as they work through it

Not generic advice.
Real paths. Real steps. Built for that specific person.

How It Works

AI Skill Matcher is a full-stack MERN application with Groq AI at its core.

1. The Quiz

The onboarding is a multi-step quiz. Users answer questions like:

What are you naturally good at?
How many hours a week can you dedicate?
What is your income goal?

These answers form the prompt sent to Groq.

2. AI Analysis with Groq

I send the quiz answers to the Groq API with a carefully structured prompt. The prompt instructs Groq to:

act as a career coach
return exactly 3 income paths
respond in clean JSON format

The response includes title, category, difficulty, income range, time to start, and required skills — all structured and ready to store. One of the biggest challenges here was making the prompt reliable. Groq sometimes wraps JSON in markdown code blocks.
So I added a cleanup step before parsing:

JSON.parse(text.replace(/```json|```/g, '').trim())

Small fix. Big impact on stability.

3. Roadmap Generation

Once a user picks a path, Groq generates a 30-day roadmap. Each step includes:

a title and description
estimated days to complete
learning resources with links

This turns a career idea into an actual plan with daily actions.

4. User Dashboard

Users can:

save their favourite paths
track which roadmap steps they have completed
come back and continue from where they left off

The dashboard is powered by Zustand for state and MongoDB for persistence.

5. Authentication

I built a full JWT auth system:

register and login
passwords hashed with bcrypt
protected routes on both frontend and backend
axios interceptor adds the token to every request automatically

Challenges

1. Prompt Reliability

Getting Groq to return consistent, parseable JSON every time was the hardest part. The solution was a combination of:

very explicit prompt instructions
JSON cleanup before parsing
fallback error handling if parsing fails

2. Protecting the AI Quota

Groq API calls cost quota.
If users spam the AI, it runs out fast. I added rate limiting specifically on the AI routes:

const aiLimiter = rateLimit({
  windowMs: 60 * 60 * 1000, // 1 hour
  max: 10                    // 10 AI calls per user per hour
});

This keeps the app stable without blocking normal usage.

3. Schema Design

The data relationships were tricky: User → QuizResponse → Path → Roadmap Getting this right upfront made the API layer much cleaner.
Bad schema design early would have caused pain later.

What I Learned

This project taught me:

how to engineer prompts that return structured, reliable outputs
how to build multi-step onboarding flows with good UX
how to design MongoDB schemas with nested references
how rate limiting protects third-party API integrations
how to connect frontend, backend, database, and AI into one clean system

Final Result

AI Skill Matcher became a platform that:

listens to what a user is actually good at
uses AI to generate real, personalized career paths
gives a full action plan — not just inspiration
tracks progress so users stay on the path

Conclusion

This project is not about showing off AI features. It's about solving a real problem. Most people know their skills.
They just don't know the next step. AI Skill Matcher gives them that next step — clearly, personally, and with a plan. That's what I wanted to build.
And I think it does exactly that.