Scoura: SaaS product built in Claude Code
AI/ML
Product Strategy
SaaS
Design Systems
Full-stack Build
Brand
Overview
Scoura is a job-fit scoring tool built on a simple reframe: instead of scoring how well you match a job, it scores how well the job matches you. Users paste in a job description, and Scoura weighs it against their profile, experience, and resumes to return a clear score, band, and breakdown, comp fit, scope, red flags, and dealbreakers included, so they know whether a listing is actually worth their time before they spend any on it.
Challenge
The idea came out of my own job search, not a market gap I went looking for. Applying to enough roles that I started losing track of duplicates. No visibility into which applications had quietly gone stale with no response. And the real cost, reading full listings end to end only to find the dealbreaker, a vague comp range, an unreasonable ask, a scope mismatch, buried in the last three lines. Every existing tool in this space, LinkedIn, Jobscan, Teal, Huntr, scores the candidate against the job. None of them ask whether the job is worth pursuing in the first place.
Scoura flips that. A weighted rubric scores skills, scope, domain, growth, and red flags, with hard caps that ceiling the score when compensation falls below a stated floor or location doesn't match, and forced-transparency rules that keep the system honest at both ends, no scores clustering in the 90s without a real reason, no bottom-tier scores without a named cause.
Approach
I started by defining the reframe, score the role, not the candidate, then built the system to back it up.
I designed and built the entire product solo, including the code. The application is built in Claude Code, a Next.js app with a token-based design system built in Figma, so every component, button, input, card, table, pulls from the same source instead of getting styled one-off as new screens got added. I designed and built the entire product solo. I built Cairn, a custom design system, in Figma and made it bidirectional, so design and code stay in sync. The app is hosted on Vercel, with Supabase for the database and auth, with an AI-powered scoring engine underneath the analysis.
Every piece of the product logic, the ghost/stale application flow, the scoring caps, user logic, the free-trial abuse prevention, the pricing tiers, was designed and stress-tested against edge cases before being built, then handed off as a precise spec, not just a rough idea, and built in Claude Code from there.
Resumes, user profiles, and tracking to ensure end to end input and seamless output
