Back to projects

Case study

AI2024

OwnCover

My product — desk-ready warranty vault for GST invoices, cover dates, and claim packs.

OwnCover is a product I conceived and built end to end (originally Warranty Vault AI). It helps homes, shops, gyms, and offices keep GST invoices, cover dates, and documents in one vault — scan or forward bills, track manufacturer vs store cover separately, get reminders, and download a printable claim pack before the service centre, without OwnCover filing the claim for you.

Live site
OwnCover marketing hero — every bill and cover date in one vault.
Next.jsTypeScriptTailwind CSSPostgreSQLGeminiNextAuth

Gallery

Contribution

  • Conceived OwnCover from a real service-desk pain point and built the live product at owncover.hvtx.in — idea, UX, marketing site, vault app, and backend.
  • Built GST QR-first and OCR capture with user-confirmed dates before anything is saved.
  • Implemented manufacturer vs store cover tracking, 30/7/1-day reminders, and calendar exports.
  • Shipped desk-ready claim pack PDFs and inbound email forwarding for Amazon, Flipkart, and Croma invoices.
  • Designed shared-vault flows so families and small businesses can share records with separate sign-ins.

Outcome

Took OwnCover from idea to a live service — a desk-focused warranty vault that turns scattered bills into a printable claim pack before visiting the service centre.

Technical decisions

  • Document-first capture flow

    I designed the product around scan, review, and confirmation so extraction could be corrected before anything hit the vault.

    Next.js · Gemini

  • Marketing site separate from vault workspace

    I kept the public story focused on desk outcomes while the app handled products, documents, reminders, and pack downloads.

    Next.js · NextAuth · PostgreSQL

  • Inbound invoice pipeline

    I built private forwarding addresses that create draft records from marketplace PDFs for confirmation in-app.

    Next.js · PostgreSQL

Challenges

  • Messy GST and marketplace invoice formats

    Invoices arrive as QR codes, photos, email PDFs, and screenshots with inconsistent fields and Hindi/English labels.

    Resolution: I used QR-first parsing with on-device OCR as fallback, left fields empty when unsure, and required explicit date confirmation before persistence.

  • Two cover clocks at the service desk

    Manufacturer and store or AMC end dates are easy to confuse when a claim is rejected for expired cover.

    Resolution: I modeled cover types separately, surfaced a timeline in the vault and claim pack, and tied reminders to the date that still matters.