Job Tracker + AI Tools
AI-powered job application tracker — Kanban pipeline with streaming cover letters, fit scoring, and interview prep
3 Claude-powered tools
SSE streaming cover letters
Visual score ring UI
8 Kanban pipeline stages
01The Problem
Job searching is itself a full-time workflow: tracking applications, following up at the right time, writing tailored cover letters for each role, and preparing for interviews. Most people manage this in spreadsheets. The opportunity was to combine standard application tracking with AI tools that are actually useful at every step of the funnel.
02The Approach
Built a full-featured Next.js app with a Kanban board (8 pipeline stages), application detail pages with timeline events, and then embedded three AI tools directly into the workflow. From any application, one tap takes you to the AI tools page with company and role pre-filled. Cover letters stream in real time. Job fit analysis returns a 0-100 score with specific strengths, gaps, and talking points.
03Architecture Decisions
Three specialized Claude API routes
Cover letter: SSE streaming to client — text chunks arrive token-by-token for a live typing effect. Job fit: structured JSON with a typed schema enforced via prompt engineering — returns overallScore, verdict, strengths array, gaps with severity, talkingPoints, redFlags, and salary context. Interview prep: 8-10 questions each with category, difficulty, whyAsked rationale, and a Harrison-specific angle tailored to his background.
URL-based navigation between tracker and AI tools
Clicking 'Cover Letter', 'Job Fit', or 'Interview Prep' on an application detail page navigates to /ai with query params (?tab=cover-letter&company=Anthropic&role=AI+Engineer). The AI page reads these via useSearchParams() and pre-fills the form. No shared state, no prop drilling — URL as state.
Harrison's full profile embedded in system context
All three routes include Harrison's detailed background — research publications, 28 shipped apps, current role, key projects. The AI generates letters that reference specific papers, specific apps, not generic filler. Personalization happens at the prompt level.
Visually communicative job fit UI
The job fit result is built around a custom SVG score ring (animated stroke-dasharray) that renders the 0-100 score visually before the user reads a word. Strengths are color-coded emerald, gaps have a severity dot system (minor/moderate/major), talking points are blue cards, red flags are rose-tinted warnings.
04Key Insight
The meta-observation: this app was in active use the same day it was built. The AI tools generate cover letters that reference specific papers, specific apps, specific GitHub activity — only possible because the AI has complete context. Most AI tools feel generic because they have no context. This one doesn't.
05Why It Matters
The most immediately practical tool in the portfolio. The combination of structured pipeline tracking + contextual AI generation tools in a single workflow is rare; most job-search tools do one or the other, not both.