LOSSVAULT
A platform that accepts bulk photo uploads and returns a structured, adjuster-ready loss list within 24 hours.
Visit Site24-Hour
SLA
1,000+
Photo Capacity
Per-Job
Pricing Model
Web + Mobile
Surfaces
The Problem
When a property is damaged by fire, flood, or storm, restoration contractors are responsible for documenting every affected item for the insurance claim. That means walking the job site, photographing everything, and building a structured loss list that the adjuster can actually work from.
The documentation process is the bottleneck. Contractors generate hundreds or thousands of photos on a single loss. Turning those photos into a clean, organized, adjuster-ready report is manual, time-consuming work that pulls them away from the actual job. For large losses it can take days.
The adjuster is waiting on that list. The homeowner is waiting. And the contractor has other jobs on the schedule. That delay is a real operational and cash flow problem, not just an inconvenience.
What We Built
LossVault is a platform that accepts bulk photo uploads and returns a structured, adjuster-ready loss list within 24 hours. Contractors upload their job site photos, pay based on volume, and get back a complete, itemized report they can hand directly to the adjuster.
The core of the platform is a proprietary analysis pipeline that processes each photo and extracts structured data: item identification, category, brand and model when visible, condition assessment, replacement cost, confidence score, and a rationale. Every result is validated against a strict schema before it hits the database, and a quality review layer ensures no output reaches the user without human sign-off.
Job Workflow
Contractors create a job, upload their photos in bulk (1,000+ in a single session), and pay via Stripe Checkout before processing begins. The system queues the images and processes them sequentially through the analysis pipeline. A progress tracker keeps the contractor informed in real time.
Before results are published, every job goes through an admin review queue. This is a deliberate quality control gate. An admin reviews the output, can retry flagged images, and approves before the contractor sees the final list. This matters because the output goes into an insurance claim. It has to be right.
Vision Analysis Pipeline
The analysis layer processes each photo as an independent line item. For every image it returns a structured record: item description, category, brand, model, quantity, condition, replacement cost, low and high price comparables, comparable sources, a confidence score between 0 and 1, and a written rationale. Items flagged as needing review are surfaced separately so the admin can inspect them before approval.
The system is built to handle ambiguity correctly. If a brand or model cannot be identified with certainty, the field returns N/A rather than a guess. If multiple similar low-cost items appear in a single photo, they are grouped and priced as a lot. Structural fixtures are excluded automatically.
Output & Export
Once approved, contractors access their results through a filterable analysis table. Every item is displayed with its valuation, condition, comp sources, and confidence score.
Jobs can be exported in two formats. The Excel export is a password-protected XLSX with three sheets: Claim Summary, Valuation Summary, and Line Item Details. The PDF export is a landscape-formatted report with automatic page breaks and page numbering. Both formats are designed to be handed directly to an adjuster without any additional formatting work.
Adjuster Sharing
Contractors can generate a tokenized share link for any completed job. The adjuster receives that link and can view the full loss list and export it without creating an account. This eliminates the friction of back-and-forth file transfers and gives adjusters direct access to the structured data.
Pricing Model
LossVault is pay-per-job. Contractors pay based on photo volume at the time of submission: under 1,000 photos for $99, 1,000 to 5,000 for $299, 5,000 to 10,000 for $499, and anything above 10,000 at $500 base plus $0.25 per additional image. Payment is processed through Stripe Checkout before the job enters the processing queue.
The per-job model was a deliberate choice. Restoration contractors do not work on a predictable monthly volume. A subscription model creates friction; pay-as-you-go matches how they actually work.
Referral Program
LossVault has a built-in referral system with tiered incentives. New users who sign up through a referral code receive 50% off their first job. The referrer earns commissions on the first three jobs their referral completes: 50% on the first, 30% on the second, 15% on the third. All rewards are tracked and processed automatically when jobs are approved and published.
Built-in CRM
The platform includes a CRM for prospecting restoration contractors. Accounts, activity timelines, follow-up scheduling, pipeline stage tracking, and CSV import are all built in. This is the internal sales tool used to grow the user base.
Mobile App
The mobile app is the consumer-facing surface, built for the job site. Contractors use it during the walkthrough itself: open a project, photograph items as they move through the property, and the system begins processing in parallel. By the time the walkthrough is done, processing is already underway.
The mobile app also includes a contents and packout management system for tracking items removed from the property, their condition, and storage location.
Architecture Decisions
The pay-before-process model is both a business and an architecture decision. It eliminates the risk of processing jobs that never convert, keeps the queue clean, and aligns the contractor's commitment with the start of work. The pricing tiers are computed server-side and validated against the actual uploaded image count before the Stripe session is created, so the amount cannot be manipulated client-side.
The admin approval gate was added deliberately. Insurance claim documentation has a higher accuracy bar than most consumer software. Rather than shipping automated output directly to the user and hoping it is correct, every job goes through a human review before the contractor sees it. This adds latency but eliminates the trust problem.
The file-hash deduplication on upload prevents the same photo from being processed twice within a job. It happens at ingest, not at processing, so it does not add overhead to the analysis pipeline.
The tokenized adjuster share system uses server-side token validation on every request. The adjuster cannot access any other job by manipulating the URL. The share page is completely decoupled from the authenticated portal routes.
Tech Stack
Frontend / Platform
Processing & Automation
Billing & Integrations
Data & Security
My Role
Co-founder and technical co-architect. Scope includes the full-stack web platform design and development, the processing pipeline and schema architecture, the Stripe billing integration, the export systems, and the referral program infrastructure. The mobile app is in active development.
SOME PROBLEMS DESERVE AN ENGINEER.
Not a template. Not a prompt. A system built to last.
Start a Project