Coming SoonProduct Founder & System Engineer

MyKar

AI Fleet & Vehicle Expense Management SaaS

Next.jsTypeScriptGPT-4o VisionPostgreSQLFastAPITailwindCSS

Problem

Small logistics fleets and rental companies struggle with manual paper fuel receipt logging, missed maintenance schedules, and untracked vehicle operating costs.

Solution

Architecting MyKar, an AI mobile-first SaaS that allows drivers to snap photos of fuel receipts for instant OCR data extraction while auto-calculating per-kilometer vehicle efficiency.

System Architecture

Next.js Progressive Web App uploading receipt images directly to S3. Vision LLM OCR extracts vendor, date, liters, and total cost, writing structured records to PostgreSQL.

System Architecture Flow

3 Components

STEP 01

Driver PWA Interface

Mobile-optimized camera upload UI.

STEP 02

Vision OCR Pipeline

GPT-4o Vision receipt extraction engine.

STEP 03

Fleet Analytics Engine

Per-kilometer cost & maintenance interval math.

Execution Workflow

1. Driver snaps photo of fuel receipt at gas station.

2. Vision OCR extracts total price, liters, and station name in < 3s.

3. Odometer reading input updates vehicle maintenance schedule.

4. Fleet manager views live operational cost analytics.

Business Impact & Metrics

Provides fleet owners with real-time operational cost transparency and automated maintenance alert schedules.

Receipt Logging Time3s

Before: 5 mins manual entryAfter: 3s AI scan

Data Extraction Accuracy98.5%

Before: Manual typosAfter: 98.5% accurate

Key Lessons Learned

Crinkled or thermal-faded fuel receipts fail traditional Tesseract OCR; multimodal Vision LLMs are vastly superior for real-world receipt parsing.

What I Would Build Differently Today

“I am building MyKar with a offline-first SQLite sync engine on the client (using PowerSync/Replicache). This ensures drivers can snap receipts deep inside underground parking structures without losing data.”