DEV Community: MADAN DHOUNDIYAL

My Stack for ResumeOrbitz: React, Supabase, Netlify, and Why

MADAN DHOUNDIYAL — Sun, 18 Oct 2026 13:00:00 +0000

One person, zero servers to babysit.

This is every tool behind ResumeOrbitz, a free AI resume builder, and the one choice I'd change if I started today.

New here? Start with the backstory: How I Built an AI Resume Builder as a Solo Founder in India.

The constraint decided everything

I'm the only engineer, designer and marketer. So every tool had to pass three tests:

Ship fast: a stable version in weeks, not months
Low budget: free tiers first, paid only when it saves real time
No ops: if it needs a server I have to patch, it's out

Every choice below falls out of those three rules.

The stack at a glance

Layer	Tool	Why	Cost / month
Frontend	React + TypeScript + Vite	Stateful editor, fast builds	Free
Styling	Tailwind CSS	Styles live next to each template	Free
Auth + database	Supabase	Postgres, auth and row-level security in one	Free tier
Hosting	Netlify	Push to deploy, no servers	Free tier
Analytics + SEO	GA4, Search Console, JSON-LD, llms.txt	Free and good enough	Free
AI	Groq API	Fast responses, free tier	Free tier
Pair programmer	Claude Pro	One assistant that holds the whole problem	Pro plan (building only)

Frontend: React + TypeScript + Vite

A resume editor is a stateful UI. You type on the left and the preview updates on the right. React's component model fits that well, and its ecosystem meant I never had to build a basic widget myself.

TypeScript earns its place because a resume is a deeply nested object: sections, then entries, then bullets. Rename one field and the compiler shows every place that breaks, before a user finds it.

Vite keeps the loop short: the dev server starts in seconds, and hot reload shows a template change the moment I save.

Styling: Tailwind CSS

With 150+ templates, CSS can turn into a mess fast. Utility classes keep each template's styles next to its markup, so changing one template doesn't quietly break another.

Tailwind also strips unused classes at build time, so 150+ templates don't mean a huge CSS file.

Backend: Supabase for auth, data and storage

Supabase gave me Postgres, auth and row-level security without writing a backend. Saving a resume is one call from the frontend:

const { error } = await supabase
  .from("resumes")
  .upsert({
    id: resume.id,
    user_id: user.id,
    data: resume,
    updated_at: new Date().toISOString(),
  });

The part that makes this safe is a row-level security policy, so a user can only ever touch their own resumes:

create policy "Users manage their own resumes"
on resumes for all
using (auth.uid() = user_id);

With that policy in place, even a bug in the frontend can't hand one user another user's resume.

Hosting: Netlify

Git push, and it's live. Every branch gets a preview URL, which is handy for testing a template before it ships.

The one file every single-page app on Netlify needs is public/_redirects:

/*    /index.html   200

Without it, refreshing any page other than the home page returns a 404.

SEO in a single-page app

Google Analytics only counts the first page load in an SPA. This small hook records every route change as a page view:

import { useEffect } from "react";
import { useLocation } from "react-router-dom";

export function usePageTracking() {
  const location = useLocation();

  useEffect(() => {
    window.gtag?.("event", "page_view", {
      page_path: location.pathname + location.search,
      page_title: document.title,
    });
  }, [location]);
}

On top of that:

A JsonLd.tsx component, wired into App.tsx, adds JSON-LD schema (including FAQPage) to each page
Open Graph tags, so shared links show a proper preview
An llms.txt file that tells AI assistants what the site does
Google Search Console from early on

AI: Groq

Groq runs the AI behind ATS scoring, keyword gap analysis and cover letters. I picked it for two reasons: responses come back fast enough to feel instant inside the editor, and its free tier covers what a solo product needs right now.

The trade-off is rate limits on the free tier. That's a good problem to have, and one I'll solve when usage forces it.

What I'd change

I'd start with pre-rendering or SSR. A single-page app hands search engines an almost empty HTML shell until JavaScript runs. For a product that lives on organic search, every SEO fix since (JSON-LD, the page-view hook, llms.txt) has partly been a workaround for that one early choice.

If I started today, landing pages would ship real HTML from day one.

What it costs per month

₹0 to run. Supabase, Netlify, Groq, GA4 and Search Console are all on free tiers. The only paid tool is my Claude Pro subscription, and that's for building, not running.

Your turn

What's your stack for solo projects, and what would you swap? I read every comment.

Disclosure: I'm the founder of ResumeOrbitz, the tool this post is about.

How I Built an AI Resume Builder as a Solo Founder in India

MADAN DHOUNDIYAL — Tue, 13 Oct 2026 12:54:22 +0000

Most resume builders let you write for free, then ask you to pay when you hit Download.

I built one that doesn't. I did it alone, from India, and the first stable version took 15 days.

This is the story of how ResumeOrbitz came together: why I built it, the hardest parts, and what I'd tell anyone starting something similar.

Why I built it

It started with my own job switch.

I was updating my resume, and every time I made changes, the platform asked me to pay before I could download it. The free ones weren't much better. Either the useful features were missing, or the good templates were locked behind "premium".

That's when it hit me: if I'm facing this, thousands of other job seekers are too.

So I decided to build an AI-driven resume app that would be completely free. Not a free trial, and not "free with a paywall at the end". Just free.

What ResumeOrbitz does

I'll keep this short, because this isn't a sales post.

ATS score: checks how well your resume will survive an applicant tracking system
Keyword gap analysis: compares your resume with a job description and shows what's missing
150+ templates, with both Indian and US formats, and none locked as "premium"
Cover letter generation from the same profile
Free: no paywall at download time

The stack, in one paragraph

The frontend is React with TypeScript, built with Vite and styled with Tailwind. Supabase handles auth and the database, and Netlify hosts it. AI features run on LLM APIs. There are no servers for me to babysit, which matters a lot when you're the only engineer, designer and marketer.

I'll break down every layer in my next post.

Hard part #1: Building a full product with no team

I'm a software engineer, so I knew how to start and roughly how the pieces should fit together. But knowing how to start is very different from building an entire custom app alone.

Three problems hit me right away:

No team. There was nobody to own the frontend, the backend or the architecture. Every decision, and every bug, was mine.
Too many layers at once. UI, database, auth, AI features, hosting and SEO all had to work together, and I had to switch between them constantly.
Free AI tools kept running out. I started by juggling free sessions on different AI models. Each one had usage limits, so I'd lose context in the middle of a problem and have to start over somewhere else.

The fix: I stopped juggling and moved everything to one paid tool, Claude Pro. It became my pair programmer, architect and debugger in one place. Having one consistent assistant that could hold the whole problem made a big difference to how fast I could move.

The lesson: when you're solo, your tools are your team. Paying for one good tool cost far less than the time I was losing by switching between free ones.

Hard part #2: Deciding what an "ATS score" means

There's no official ATS standard. Every company uses different software, and none of them publish how they rank resumes. So any ATS score is an estimate, and I wanted mine to be an honest one.

So I built the checker around the job description. You paste in the JD from Naukri, LinkedIn or any portal, and the AI scans your resume against it for the things that reliably cause trouble:

Keyword match: does your resume use the exact words the JD uses? "NodeJS" and "Node.js" aren't the same to an ATS.
Missing skills: which key skills from the JD aren't on your resume at all
Formatting traps: tables, columns, images and other elements that scramble ATS parsing

The score matters less than what you do with it. A number alone doesn't help anyone. So after the score, there's a one-click AI fix that rewrites or strengthens sections to close the gaps. Then you check again.

Hard part #3: SEO for a single-page app

This one surprised me. I built ResumeOrbitz as a single-page app, which is great for users but harder for search engines.

Three things made the biggest difference:

Tracking route changes. Google Analytics only counts the first page load in an SPA by default. I added a small custom hook so every route change is recorded as a page view.
Structured data. A small component adds JSON-LD schema (including FAQ schema) to each page, so search engines understand what the page is about.
An llms.txt file. This is a newer idea: a plain-text file that tells AI assistants what the site is. More people now ask AI assistants for tool recommendations, so it felt worth the 20 minutes.

I also set up Google Search Console early. If I started again, I'd do all of this before launch, not after.

The rename: ResumeForge to ResumeOrbitz

The product started life as ResumeForge. Then I found out the name and the domain were already taken.

Renaming mid-build was annoying: new domain, new logo, new repo name, and every reference in the codebase. But it was much cheaper to do it then than after launch, and I ended up liking ResumeOrbitz more anyway.

Tip: check the domain, the social handles and a quick trademark search before you design a logo.

The solo founder timeline

Here's how it actually went:

15 days to build the first stable version
Then small enhancements, one at a time, based on what I noticed and what users needed
Now, 99% of my time goes into marketing

That last point is the one nobody warned me about. Building turned out to be the easier half. Getting the product in front of the people who need it is the real work.

What I'd tell someone starting out:

Pick managed tools. Every hour spent on infrastructure is an hour not spent on users.
Ship before it feels ready. My first stable version took 15 days, not 6 months, and that was the right call.
Plan marketing from day one. You'll spend far more time on it than you expect.

Where it stands now

ResumeOrbitz is live and completely free. I've paused big new features, and almost all my time goes into reaching job seekers who need it. This blog is part of that.

What's next

Next I'm writing up the full stack: every tool, why I picked it, and what I'd change.

A question for you: what would you want from a free resume tool that the current ones don't give you? I'll read every answer.

Disclosure: I'm the founder of ResumeOrbitz, the tool this post is about.