I got tired of manually reading JDs, so I built a tool to do it for me

Honestly, this started out of frustration.

I was applying for jobs a few months back and kept doing this thing where I'd open a job description, read through it, go back to my resume, read through that, and try to mentally map which keywords I was missing. Every. Single. Application.

After doing this about a dozen times, I thought — This is just text matching. Why am I doing this by hand?

So I built CV Craftery. You paste your resume and a job description, it extracts the relevant keywords, scores how well your resume matches them, and shows you exactly what's missing. There's also an optional AI rewrite of your summary and skills section if you want it to fill in the gaps — but honestly, the scoring alone is what I use most.

The stack

Nothing exotic — Spring Boot on the backend, hosted on a t2.micro EC2 instance behind Nginx with SSL from Certbot. For the AI part, I'm using Groq as the primary, with Gemini and Anthropic Claude as fallbacks. All three have rate limits on free tiers, so I ended up building a mechanical keyword injection fallback too — just in case everything else gives up. Inelegant, but it means the core feature never just... fails.

The thing that caught me off guard

Spring Boot was generating http:// redirect URLs even though the site is HTTPS.

Took me longer than I'd like to admit to figure out what was happening. Nginx proxies to localhost:8080 internally, so Spring Boot sees the connection as plain HTTP and builds redirect URLs accordingly. The fix was one line:

server.forward-headers-strategy=native

That tells Spring Boot to trust the X-Forwarded-Proto header that Nginx is already sending. Obvious in hindsight, but I went down a few wrong paths first.

What I'd do differently

I'd think harder about the domain name earlier. I'm using a subdomain of an existing domain I had access to (cvcraftery.amsorg.co.in), and it's caused some headaches — a couple of AI tools flagged it as suspicious just because it's a multi-level subdomain with no trust history. The site is totally legitimate, but the domain pattern looks sketchy to automated scanners—nothing I can do about it now, but worth knowing if you're starting something new.

Where it's at

It works. No account needed, no paywall, files deleted within an hour. I've been using it myself, and it's genuinely saved me time.

If you try it, I'd really appreciate honest feedback — especially if you work in recruiting or have used actual ATS systems and can tell me whether what I'm extracting actually matches what those systems weigh. That's the part I'm least confident about.

Link: https://cvcraftery.amsorg.co.in