This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend
What I Built
My friend Vanshika has been stuck on one question: which job should I even aim for? The usual advice is "just do LeetCode", and that doesn't help when you don't know which path fits you in the first place.
I kept thinking there had to be a better way than a personality quiz or a random YouTube roadmap. So I built Trailhead, a placement guide where you try the path before you pick it.
Here's how it works:
- It reads what you've actually built. You give it your GitHub username (and optionally paste your resume). It looks at your repos and works out which skills you can really prove.
- It scores 9 career paths like web, backend, QA, data, DevOps, AI apps and mobile. Every score comes with the evidence behind it, so you can see why.
- It lets you taste-test. For each of your top paths there are three quick questions. Then you rate how much you enjoyed it, because liking something matters as much as being good at it.
- It builds a 30-day plan that fits your free time and uses the stack you already know. If you miss a few days, it re-plans instead of making you feel behind.
What Vanshika said when I showed her:
"This app really helped me decide which path to take. It looked at my GitHub and gave me my path: where I'm heading and how I should work towards it."
Demo
Live app: https://trailhead-k0vy.onrender.com
Don't want to type anything? Click Try an example. It walky for a made-up student, no sign-up. (It runs on a free server, so the very first load can take a few seconds to wake up.)
Code
yashsrivastava1408
/
TrailHead
An evidence-based placement guide for final-year students who don't want to grind DSA and don't know which career path fits them.
Trailhead
Try the path before you pick it.
An evidence-based placement guide for final-year students who don't want to grind DSA and don't know which career path fits them.
Contents
- The problem
- What Trailhead does
- Screenshots
- Design principles
- Architecture
- Tech stack
- Quick start
- Configuration
- Scripts
- Project structure
- Testing
- Deployment
- Production readiness
- Security and privacy
- Limits and roadmap
- Documentation
The problem
Final-year students are told two things: "start LeetCode" and "pick a role". Many people hear that and freeze.
| # | What goes wrong today | Why it stays unsolved |
|---|---|---|
| 1 | They don't know which path fits them. Web, QA, data, DevOps, support, AI apps, mobile: the options are blurry. | Career quizzes ask what you say you like, and people answer them badly. |
| 2 | "Just do DSA" is the only advice they get. | Generic advice ignores what the student has already built. |
| 3 | They can't tell what they would enjoy until they try |
How I Built It
It's React and Express, with LangGraph.js, SQLite, and an open-weight model (gpt-oss-120b) running on Groq's free tier.
The rule I kept coming back to: code does the counting, the model only writes the words, and something checks the words.
- Skills, scores, grades and re-planning are plain, predictable code. No guessing.
- The model writes the "here's why this path fits you" text, bint at evidence from your own profile. A small checker rejectsanything it can't prove and sends the model back to try again. That loop is a tiny LangGraph graph.
- If the model is down or out of quota, you still get real scoand the screen tells you why. The taste tests never use themodel at all.
The part I'm most glad about is what testing caught:
- A resume line saying "I'd rather avoid DSA" was being read as DSA experience. The DSA-heavy path scored 58% for someone who avoids DSA. Now it's 17%.
- The quiz was sending the correct answers to the browser. Oops.
- Everyday words were creating fake skills. "Time to go home"
- It was storing resumes that nothing needed. Now it never does.
There are 113 server tests, 94 client tests, and a live suite that runs against the real model and the real GitHub API. I also clicked through the
whole thing in a real browser to make sure it behaves.
What I'd be honest about: it only sees public GitHub worpublic code gets a "not enough evidence" warning instead of aconfident guess. The taste test is three questions per path, a quick signal and not an exam. And it's one small server with one database file, which is
right for this, not for a million users.
Why Does Open Innovation Matter?
Honestly, the clearest example happened to me. Llama wasn't enWith a closed API I'd have been stuck with whatever I washanded. Because the model is just a setting, I pointed Trailhead at a different open-weight model by changing one line. Nothing else in the code changed.
It also matters for privacy. A resume is personal, so Trailheaause the model sits behind one setting, a college could runthis against a model on its own machines.
And it costs nothing to run, which matters when the people using it are students.
One honest note: my demo calls a hosted API, so it is not offline. Running a local model works with the same code path, but I haven't tested that
yet, so I won't claim it.
Prize Categories
- Best Use of Render: the app is deployed on Render as one
Quick decisions
If the quote was actually your own words, swap the whole "Whathis:
What happened when I ran it on my own GitHub:
"It really helped me decide which path to take. It looked atpath: where I'm heading and how I should work towards it."
Top comments (0)