A multi-tenant cloud surveillance suite spanning a Web Dashboard, Super Admin Panel, Mobile companion, an auto-scaling Streaming Orchestrator microservice, and a Cloudflare Worker edge proxy โ delivering real-time AI event detection, multi-channel alerting, and on-demand analytics over RTSP and device cameras.

Designed the five-component architecture โ Web Dashboard, Super Admin Panel, Mobile companion, Streaming Orchestrator microservice, and Cloudflare Worker HLS edge proxy.
Built the React Web Dashboard with multi-camera live grid (up to 15-camera layouts), pull-based stream lifecycle, low-latency mode, and the AI event polling feed.
Implemented camera registration for RTSP/IP cameras with FPS control, zone definitions, health monitoring, and RTSP URL caching for faster stream startup.
Engineered the Streaming Orchestrator FastAPI microservice with MediaMTX lifecycle management, auto-scaling DigitalOcean droplet provisioning, buffer pools, and stream rebalancing.
Shipped the Cloudflare Worker edge proxy for HLS playlist and segment caching with proper CORS for cross-origin browser playback.
Delivered all 8 AI event types (person, loitering, intrusion, motion, vehicle, PPE, safety, zone activity) with bounding-box snapshot annotation and severity scoring.
Built the multi-channel alerting system across email, SMS, WhatsApp via Twilio, plus Slack, Teams, Zapier, Make.com, and n8n webhook integrations.
Implemented the Stripe subscription engine โ five plan tiers, hosted checkout, billing portal, webhooks, trial management, and camera-limit enforcement with upsell prompts.
Created the analytics suite covering people, vehicle, security risk assessment, camera and zone metrics, with on-demand report generation.
Delivered the mobile companion experience โ getUserMedia device camera streaming, QR code pairing, mobile-optimized live grid, and on-site alert triage.
Added internationalization with English and Hebrew including full RTL layout support across the customer-facing surfaces.
A complete cloud surveillance pipeline from RTSP camera ingest to multi-channel alerts, with real-time AI event detection, auto-scaling streaming infrastructure, and edge-cached low-latency playback.
๊ฐ์ ์์ค์ ์ ๋ฌธ์ฑ๊ณผ ํ์ ์ผ๋ก ๊ทํ์ ๋น์ ์ ์คํํ๋ ๋ฐฉ๋ฒ์ ๋ํด ๋ ผ์ํด ๋ณด๊ฒ ์ต๋๋ค.

MicrocosmWorks designed AwareCam with a pull-based stream lifecycle that only activates streaming when a viewer requests it, with TTL-based automatic shutdown. AI event detection runs on the cloud side where the auto-scaling MediaMTX orchestrator processes streams, analyzing frames for 8 event types including person detection, loitering, and zone intrusion. The Cloudflare Worker edge proxy caches HLS segments to reduce origin bandwidth, while bounding-box annotation snapshots are stored rather than full video, minimizing storage costs.
MicrocosmWorks delivered 8 AI event types: person detection, loitering, intrusion zone violation, motion detection, vehicle detection, PPE compliance, safety hazard, and zone activity monitoring. Each event includes bounding-box snapshot annotation and severity scoring that feeds into the alerting engine. The system supports configurable detection sensitivity thresholds and zone definitions per camera, allowing operators to reduce false positives in their specific environment.
MicrocosmWorks built an event-driven alerting pipeline that routes AI detections to configured channels based on event type and severity. Email and SMS alerts go through Twilio, WhatsApp notifications use the Twilio WhatsApp Business API, and Slack and Microsoft Teams receive webhook-formatted alerts. Additionally, Zapier, Make.com, and n8n webhook integrations allow operators to connect AwareCam alerts to any downstream workflow system.
MicrocosmWorks engineered the Streaming Orchestrator as a FastAPI microservice that manages MediaMTX server lifecycle with auto-scaling DigitalOcean droplet provisioning. When camera load exceeds a server's capacity, the orchestrator spins up additional streaming nodes, distributes cameras across the pool using buffer-based load balancing, and rebalances streams when nodes are added or removed. The entire provisioning and scaling process is automatic and transparent to end users.
MicrocosmWorks implemented 5-tier Stripe subscription billing with camera-limit enforcement at the API level. Each plan tier permits a specific maximum number of registered cameras, and the system blocks camera registration attempts that would exceed the limit while showing upgrade prompts. The subscription engine handles hosted checkout, billing portal access, webhook-driven plan changes, trial management, and organization-level billing for multi-tenant deployments.