Edge AI, cloud dashboards
YOLO11 + ByteTrack run on-site — footage never leaves the building — while counts sync to a central platform for live occupancy, trends and alerts.
Case study — AI computer vision SaaS
IBSAR
Give every camera the power to understand.
- Role
- Co-founding engineering, platform architecture
- Timeline
- 2024 — ongoing
- Stack
- Angular 18 · .NET 9 · Python / YOLO11 · PostgreSQL · InfluxDB · Qdrant
RECORDED FROM THE RUNNING PRODUCT
Overview
IBSAR turns ordinary CCTV cameras into intelligent sensors — people counting, face-recognition attendance and vehicle analytics — with edge AI that keeps footage on-site and a multi-tenant cloud for dashboards, billing and fleet management. Live in production at ibsar.ai.
Visit live site ↗99.5%
counting accuracy
30+
FPS on edge hardware
6
cooperating microservices
24/7
production uptime
What made it hard — and good
Three modules in production
People counting with directional line-crossing, ArcFace-based attendance with vector search, and vehicle & parking analytics with slot-level occupancy.
True multi-tenant SaaS
Per-client module enablement, role-based access, subscriptions and billing through a regional payment gateway — built as ~6 cooperating microservices.
Deployed for real customers
Running for education and enterprise clients across Palestine, from university campuses to retail groups.
From the product
Next case study
AcademiX