AI Applyd went live on Product Hunt yesterday. Self-hunted, sole maker, four upvotes and one comment as of right now. Quiet launch, no fanfare. Posting on dev.to because devs read for the stack, not the hype.
What AI Applyd does
You paste a job URL. It scores your resume against the JD's ATS keyword extraction (deterministic, not LLM grading), tells you which keywords are missing, rewrites the resume to fix the gaps, then a headless browser agent submits the application on Greenhouse, Lever, Workday, Ashby, iCIMS, and LinkedIn Easy Apply.
Free tier covers one full auto-apply per month, plus 10 ATS scores and 5 resume tailors. No credit card.
The stack (since this is dev.to)
- Frontend: TanStack Start + TanStack Router + TanStack Query + Tailwind + shadcn/ui
- Backend: Hono on Cloudflare Workers, OpenAPI baked in
- Database: Drizzle ORM + Cloudflare D1 (SQLite at the edge)
- Storage: Cloudflare R2 for resumes, KV for caches
- AI gateway: OpenRouter for every model (paid Gemini for user-facing, free tiers for system paths)
- Browser: Browserbase + Stagehand for the auto-apply agent
- Auth: Better Auth
- Analytics: PostHog
- Payments: Stripe
- Monorepo: Turbo + Bun
- Email: React Email + Resend
Everything runs on free or near-free tiers except Browserbase (and only on user-triggered apply runs).
What is still rough
Custom (non-ATS) career pages trip the agent up sometimes on certain custom flows. We detect the failure, return the unused token reservation, and surface a "we couldn't complete this, apply manually" message. Faking a success would be worse than admitting it.
Why I built it
I applied to 127 jobs and heard back from 3. ATS filters auto-reject most resumes before any human reads them. Someone needed to do the boring part.
Link
https://www.producthunt.com/posts/ai-applyd-2
If you want to roast the stack choices, the auto-apply UX, or the pricing math, reply here. I will respond to everything today.
Top comments (0)