PrismCV

I built PrismCV to manage one canonical resume, create job-specific versions, and track applications in one place. It combines a versioned resume diagnostic with Claude-assisted tailoring and collects jobs from seven sources: Remotive, RemoteOK, We Work Remotely, Himalayas, BuiltIn, and Greenhouse and Lever job boards.

• Solo product operation

  I manage the architecture, billing, infrastructure, support, and releases.

• Assisted resume work

  Claude supports tailoring and scoring.

• Subscription billing

  Stripe checkout and webhook handling with deduplication and retries.

• Job aggregation

  Seven job sources feed a deduplicated catalog scored for each user.

A multi-version resume system that treats every tailored variant as a branch off a canonical profile:

• Branch & TailorCreate job-specific variants of a single source resume without losing the original
• AI TailoringClaude rewrites bullets and summaries against a pasted job description in one click
• LinkedIn ImportImport a LinkedIn data export ZIP into the structured editor
• Document Parsingmammoth and unpdf import existing DOCX and PDF resumes into the structured editor
• PDF ExportGenerates ATS-friendly PDFs with deterministic layout

A scoring engine plus a backend job pipeline so users can score-and-apply in one place:

• Versioned resume diagnosticChecks return explicit, fixable feedback; the public methodology explains the current evaluation model.
• Keyword MatchDiff a resume against a target job description and surface missing keywords
• Seven-source catalogScheduled jobs collect listings from Remotive, RemoteOK, We Work Remotely, Himalayas, BuiltIn, Greenhouse, and Lever
• Personalized MatchEach user sees jobs scored against their own keywords, location, and salary preferences
• Kanban TrackerDrag applications through Applied / Phone Screen / Interview / Offer with full status history

The product includes billing, account access, email, and browser-capture operations:

• Stripe billingPaid plans with idempotent webhook handling and a deduplication table
• Daily Job DigestResend-powered email digest of newly-matched jobs scored against user preferences
• Browser ExtensionChrome/Firefox extension captures jobs from arbitrary sites into the catalog

The app and database run on a self-managed Hetzner VPS.

Frontend

Next.js 16 App Router, React 19, MUI 9, MUI X DataGrid

Backend

200+ Next.js API routes for the web application

Database

PostgreSQL with Prisma 7 (50+ models, automated migrations on deploy)

AI

Anthropic Claude API for tailoring, cover letters, and job-resume scoring

Rate limiting

Redis-backed fair-use throttling

Auth

NextAuth with Google OAuth, credentials, and TOTP 2FA via otplib

Payments

Stripe subscriptions with webhook idempotency

Email

Resend for transactional and digest emails

Observability

Sentry for errors, server-side analytics in Postgres

Hosting

Self-managed Hetzner VPS via Coolify, auto-deploy from staging and main

• Prioritize as a solo operatorOwning product, design, engineering, billing, support, and infrastructure keeps the feature list tied to customer use.
• Share authorization across product surfacesThe dashboard, billing, and account actions use one authorization model so product operations do not create a separate access path.
• Know the self-hosting tradeoffThe Next.js app and Postgres share one Hetzner VPS. This keeps costs predictable and leaves maintenance with me.
• Treat model calls as an external dependencyTailoring and scoring have separate retries, prompt versions, and cost controls.