What Changed
A manual, time-intensive process of searching through security footage was replaced with an automated system that matches videos to invoices using AI.
Where They Started
Oil Changers operates security cameras across all service locations. When disputes arose or audits were required, managers had to manually search through hours of video footage to find the relevant segment tied to a specific invoice.
What Was Breaking
The manual process was inefficient and difficult to scale. Locating relevant footage required significant time and effort, and the lack of automation limited the ability to quickly resolve disputes or perform consistent operational reviews.
How The Zig Fixed It
The Zig developed a multi-component system that automates video ingestion, processing, and matching.
Security footage is ingested from Google Drive, processed into frames, and analyzed using a custom-built AI inference service written in C++ that leverages a vision-language model to detect vehicles and match them to invoices.
A backend orchestration layer manages matching workflows, while a web portal provides both video-centric and invoice-centric experiences for users.
Additional features such as global search, automated linking of related footage, and secure video sharing with OTP-protected access were implemented to enhance usability.
What It Unlocked
Managers can now instantly locate relevant footage for any invoice, dramatically reducing the time required for dispute resolution and operational audits.
The system also enables scalable processing of large volumes of video data, providing consistent and reliable matching across locations.
Where the Investment Went
Investment was focused on building the AI inference engine, video ingestion pipeline, backend orchestration systems, and frontend experience. Significant effort was also dedicated to optimizing performance and ensuring reliability across distributed components.
What This Taught Us
AI systems deliver the most value when tightly integrated into operational workflows.
This project reinforced the importance of designing end-to-end pipelines—from data ingestion to user interface—that make complex AI capabilities accessible and actionable.