FairScope

• 

  Thumbmail

• 

• 

Inspiration

Homeowners often have to approve expensive repair work while knowing less about the job than everyone else in the transaction. A photo can reveal useful context, but it cannot safely expose every hidden condition. We built FairScope to give homeowners a practical second opinion before the first quote without pretending that AI can replace an inspection.

What it does

FairScope lets a homeowner upload repair photos, choose a repair category, and enter a ZIP code. It returns: a plain-English scope of work; a transparent regional price range for lower-risk work; likely cost drivers and red flags; questions to ask a contractor; and an in-person inspection recommendation when the job is risky, unclear, or unsafe to price from photos. The key product rule is simple: AI scopes the job; deterministic code calculates the money. Gemini never supplies a dollar amount. A versioned pricing engine converts structured labor and material inputs into the displayed range, while safety gates suppress photo-only pricing for electrical panels, gas lines, structural damage, water intrusion, mold, roof leaks, foundation concerns, major plumbing or HVAC work, unclear jobs, and permit-sensitive work.

How we built it

The responsive Next.js App Router experience was designed with v0 and deployed on Vercel. Repair photos go directly from the browser to a private Vercel Blob store using a short-lived, session-scoped upload token. A signed HttpOnly anonymous session owns each report, and private photo reads are authorized against report ownership before any image is streamed. Gemini on Google Vertex AI performs multimodal image understanding and returns a validated structured scope. The deterministic pricing engine then calculates the range. Amazon Aurora PostgreSQL Serverless v2 is the primary database: it persists report ownership, structured model output, calculated prices, safety decisions, and private Blob pathnames. Vercel uses OIDC and narrowly scoped cloud identities instead of long-lived AWS or Google keys.

Challenges we ran into

The hardest part was not generating an answer. It was designing a trustworthy boundary around that answer. We separated probabilistic image interpretation from deterministic pricing, added explicit inspection-only categories, kept repair photos private, and built owner-scoped persistence without forcing a user to create an account. We also handled the operational reality of Aurora Serverless resuming after idle periods while keeping the interface honest about the wait.

Accomplishments that we're proud of

Shipped a real end-to-end production workflow on Vercel. Persisted owner-scoped reports in Amazon Aurora PostgreSQL. Kept uploaded repair photos private and inaccessible by public URL. Prevented the AI model from producing or changing dollar amounts. Added safety gates that recommend an inspection instead of inventing certainty. Used short-lived OIDC identities rather than long-lived cloud credentials.

What we learned

AI is most valuable here as a bounded perception and explanation layer. Money, authorization, retention, and safety decisions need deterministic application logic and durable records. Aurora gives FairScope the relational foundation to evolve from a report into a real transaction system without rebuilding the data layer.

What's next for FairScope

The next version turns every FairScope report into an optional job request. A homeowner will be able to attach the quote they received, invite or match with verified contractors, compare bids against the same scope, and choose a service date in the app. Qualified contractors will be able to accept nearby jobs, schedule work, message the homeowner, and update job milestones. The business can monetize completed jobs with a marketplace fee while keeping the independent homeowner report as the trust-building entry point. Longer term, FairScope will add contractor verification, calendars, in-app messaging, payments, ratings, dispute support, quote normalization, and calibrated local pricing data. The report is the MVP; the destination is a trusted dispatch marketplace for home repairs.

Built With

• amazon-aurora-postgresql-serverless-v2
• gemini
• google-vertex-ai
• next.js
• react
• tailwind
• typescript
• v0
• vercel
• vercel-blob
• vercel-oidc
• zod