JobHunt

Problem

Job hunting in 2026 is still weirdly manual. You paste the same resume into fifty applications, write cover letters that sound like everyone else's, and track everything in a spreadsheet that falls apart by week two. The process punishes people who apply thoughtfully and rewards people who apply at volume. I wanted to change that equation.

My Contribution

I designed and built the entire platform — product decisions, architecture, implementation, and prompt engineering. The core features:

• JD analysis pipeline — paste any job posting, the AI extracts the company, role, and must-have requirements automatically
• Document generation — tailored resume and cover letter per application, grounded in the user's uploaded base resume, not invented
• Application tracker — Kanban and list views so users can move applications from interested through to offer without losing track
• Keyword matching — the JD analysis surface the specific keywords so users know exactly where their resume needs to land

Architecture

Next.js App Router with TypeScript, Vercel for deployment. The AI pipeline uses an LLM via API — the prompt engineering is the hardest technical problem, not the integration. Getting the model to produce output that sounds like the candidate wrote it (not like a generic AI cover letter) required multiple iterations of context injection, tone anchoring, and constraint prompting.

Server Components handle static pages and initial renders. The tracker, document editor, and JD analysis are client islands because they need local state that changes without a page reload.

Outcomes

The core loop works. A small group of people have been using it actively — applications are being tracked, documents are going out. It's early, and the numbers reflect that honestly. The signal I care about is whether people are coming back after their first use, and so far they are.

Learnings

AI output quality is the hardest problem, not the AI integration itself. The difference between "it works" and "it's actually useful" lives entirely in the prompt. I've spent more time on prompt engineering than on any other part of the stack, and that surprised me.