Back to ProjectsDeployed — User Testing

Project Case Study

LifePilot AI

AI-Powered Personal Life Admin Assistant

An AI-powered mobile app that turns personal documents, screenshots, and forms into clear summaries, extracted details, tasks, and reminders. Users scan or upload a document, OCR extracts text, AI identifies what matters, and the user reviews suggestions before anything is saved.

React Native / ExpoNestJSPostgreSQL / PrismaAWS S3OCR + AI Structured OutputRedis / BullMQFirebase Cloud Messaging
Watch DemoView Case Study
LifePilot AI home dashboard showing today's focus, upcoming deadlines, and quick actions for scanning or uploading documents

Product Overview

LifePilot AI is a mobile-first product that helps users manage personal documents and deadlines without manually reading, extracting, and entering every detail. Instead of only storing documents, the app uses OCR and AI to understand them — generating summaries, extracting important dates and amounts, and suggesting tasks and reminders that the user can review and approve before they become active.

Problem

  • Important information is often trapped inside documents — bills, receipts, rent agreements, medical slips, school forms, government documents, and screenshots.
  • Users must manually read, interpret, extract, and remember key details like dates, amounts, and required actions.
  • Reminders must be created manually, which is time-consuming and error-prone.
  • Documents are scattered across phone gallery, email, paper, and messaging apps.
  • Users forget deadlines, delay action because they have not read a document, and feel mentally overloaded by scattered responsibilities.

Target Users

Busy Adult / Personal Life Organizer

Manages personal responsibilities across work, home, family, payments, appointments, and renewals. Receives documents containing important information but may forget to act on them.

Student

Manages school forms, assignment deadlines, exam schedules, fee payments, and registration requirements with deadlines scattered across documents and screenshots.

Freelancer / Small Business Operator

Manages invoices, client agreements, receipts, payment dates, and deliverables where missing a deadline can affect business relationships.

Parent / Caregiver

Manages family-related documents, school notices, medical appointments, household bills, and activity schedules where multiple people's deadlines may overlap.

Core Document-to-Action Workflow

The product follows a focused document-to-action pipeline. Each step is designed to reduce manual effort while keeping the user in control:

Step 1

Scan or upload a document

Step 2

OCR extracts text from the file

Step 3

AI generates summary and extracts key details

Step 4

User reviews AI suggestions

Step 5

Approved tasks and reminders appear on dashboard

Step 6

Notifications fire before deadlines

Product Walkthrough

The mobile app built with React Native and Expo. Users start from the home dashboard, scan or upload a document, and follow the processing pipeline through to approved tasks and reminders.

LifePilot AI home dashboard showing today's focus tasks, upcoming deadlines, needs-review documents, and quick action buttons for scanning or uploading documents

Home Dashboard

The user's command center showing Today's Focus, upcoming deadlines, documents needing review, and quick actions to scan or upload.

LifePilot AI add document screen showing options for camera photo, image upload, PDF upload, and screenshot upload

Add a Document

Upload options including camera photo, image upload, PDF upload, and screenshot upload.

LifePilot AI upload confirmation screen showing a document that has been uploaded successfully and is being processed

Upload Confirmation

Document uploaded successfully and queued for OCR and AI processing.

LifePilot AI document vault showing a list of saved documents with titles, categories, processing status badges, and summary previews

Document Vault

All saved documents organized with titles, categories, upload dates, processing status, and summary previews.

AI Result Review and User Control

The AI Result Review screen is one of the most important parts of LifePilot AI. It protects user trust by ensuring that AI-generated summaries, extracted details, tasks, and reminders do not become active automatically. AI output is treated as a draft that the user can review, edit, approve, or reject.

LifePilot AI document summary screen showing an AI-generated plain-language summary with important dates, amounts, and suggested next steps

AI Document Summary

A concise plain-language summary identifying the document's purpose, important dates, amounts, names, and required actions.

LifePilot AI review screen showing AI-generated task and reminder suggestions with approve, edit, and reject options awaiting user approval

Review AI Suggestions

The review screen where users can view, edit, approve, or reject AI-generated tasks and reminders before anything is saved.

LifePilot AI extracted details screen showing structured information including document type, deadline date, amounts, and names extracted from the document

Extracted Details and Deadline

Structured details extracted from the document including dates, deadlines, amounts, names, and organizations — each editable before approval.

Document Vault

Uploaded documents, AI summaries, extracted details, and related tasks are saved in the user's private document vault. Each entry shows the document title, category, upload date, processing status, and a summary preview. Users can open document details, view AI output, and delete documents.

Tasks, Reminders and Notifications

Approved tasks appear in the task list with pending, completed, and overdue statuses. Reminders support configurable timing options including at deadline, one day before, three days before, and one week before. Push notifications are delivered via Firebase Cloud Messaging and deep-link to the related task, reminder, or document.

LifePilot AI tasks and reminders screen showing approved tasks with due dates, priority badges, and reminder timing options

Tasks and Reminders

Approved tasks and reminders with due dates, priority levels, and links back to the source document.

LifePilot AI Today's Focus dashboard showing the most relevant tasks and reminders due today, with upcoming deadlines and quick actions

Today's Focus

The dashboard's Today's Focus section showing the most relevant tasks and reminders due today after processing documents.

Architecture

Mobile App

React NativeExpoTypeScript

Backend API

NestJSREST APIsJWT Access/Refresh

Database

PostgreSQLPrisma ORM

Media Storage

AWS S3 Private BucketPresigned UploadsSigned Read URLs

AI Processing

OCR Text ExtractionAI Structured OutputConfidence Indicators

Background Jobs

RedisBullMQFirebase Cloud Messaging

System Diagram

  Mobile App (React Native / Expo)
      │
      ├──▶ NestJS REST API ──▶ PostgreSQL (Prisma ORM)
      │         │
      │         ├──▶ AWS S3 (presigned uploads, signed reads)
      │         ├──▶ OCR Service ──▶ AI Structured Output
      │         └──▶ Redis / BullMQ (background jobs)
      │
      Notifications: Firebase Cloud Messaging

  Auth: JWT access/refresh tokens
  AI: Structured output with confidence indicators
  User Approval: AI suggestions require explicit user review

OCR and AI Processing Pipeline

After a document is uploaded, OCR extracts text from the file. The extracted text is stored separately from the original file. If OCR output is poor or incomplete, the document is flagged for user review. Once text is extracted, AI generates a concise summary, extracts structured details (dates, deadlines, amounts, names, organizations, contact details, addresses), and suggests tasks and reminders — all returned as structured output suitable for backend validation and user review.

Private AWS S3 Storage

Uploaded files are stored in a private AWS S3 bucket. The backend controls upload and access permissions through presigned URLs. Users can only access S3 files tied to their own account. Signed URLs for media access expire automatically. No files are exposed publicly.

Background Jobs

OCR processing, AI summary generation, detail extraction, and notification scheduling are handled as background jobs using Redis and BullMQ. This keeps the mobile app responsive while processing continues on the backend. Users can leave the processing screen and are notified when results are ready for review.

Security and Privacy Decisions

JWT authentication with access and refresh tokens
AWS S3 private bucket — no public media URLs
Presigned uploads and signed read URLs that expire automatically
Users can only access their own documents and data
AI output is treated as draft — nothing is saved without user approval
Sensitive document types handled with cautious language
Account deletion supported where implemented

AI Safety and Confidence-Aware UX

AI-generated tasks and reminders never become active automatically — user approval is required
AI indicates low confidence when the document is unclear
AI must not fabricate missing dates, amounts, names, or obligations
Each AI suggestion has its own review status: pending, approved, edited, rejected, or dismissed
Warnings displayed for sensitive document types (legal, medical, financial, government)
Users can edit, approve, or reject each suggestion individually

Key Engineering Challenges

  • 1Designing an AI output format that is structured enough for backend validation but flexible enough to handle diverse document types — from bills to medical slips to government forms.
  • 2Building a review flow that gives users meaningful control without overwhelming them — each suggestion is individually editable, approvable, or rejectable.
  • 3Handling OCR quality variance gracefully — flagging poor extraction, allowing retries, and marking documents for user review when text is incomplete.
  • 4Maintaining a calm, user-controlled experience where AI is a helpful assistant — not an autonomous agent that acts without permission.

Current Product Status

LifePilot AI is fully developed and deployed, currently being tested by real users. The complete document-to-action workflow is implemented and demonstrated across the screenshots above — from scanning and uploading through OCR, AI summary generation, detail extraction, user review, and approved tasks and reminders on the dashboard.

Document scan and upload via camera, image, PDF, or screenshot
OCR text extraction with processing status
AI document summary generation
AI key detail extraction (dates, deadlines, amounts, names, organizations)
AI-generated task and reminder suggestions with user approval flow
Document vault with categories, status, and summary previews
Task management with pending, completed, and overdue statuses
Reminder management with configurable timing options
Home dashboard with Today's Focus, upcoming deadlines, and needs-review
Private AWS S3 document storage with presigned uploads
Push notifications for approved reminders

Future Roadmap

The following features are described in the product requirements document as post-MVP or out of scope for the first version. They are not yet implemented.

  • Email integration for automatic document import
  • Calendar sync with external calendars
  • Family or team collaboration mode
  • AI chat with saved documents
  • Voice notes as document input
  • WhatsApp screenshot intelligence
  • Local language support
  • Subscription billing
  • Advanced semantic search across documents
  • Recurring reminders
  • AI-generated weekly summary

Demo Video

LifePilot AI App Walkthrough

LifePilot AI — real app walkthrough

Let's Build Together

Interested in AI-native products like LifePilot?

I'm open to full-time roles, contract work, and founder collaborations in AI product engineering and full-stack development.

Contact MeView More ProjectsLinkedIn