Experience
Overview
Enterprise-scale edge computer vision and analytics platform operating across 200+ deployment sites, utilizing Angular, Node.js, React, and Python, alongside localized IoT devices and C# Windows services.
Responsibilities
Engineered multi-site operations management portals, custom Kibana analytics plugins (React, EUI, ApexCharts), and automated edge video capture bots (C#, EmguCV). Optimized large-scale Elasticsearch data pipelines and built local Python/Docker IoT gate controllers.
Challenges
- Visualizing and querying high-velocity streaming computer vision event logs efficiently across billions of records.
- Maintaining reliable, continuous RTSP stream capture from on-premise ONVIF cameras with intermittent internet connections.
- Reducing false claims on shipping returns without introducing manual oversight delays.
- Creating a resilient system to execute scheduled tasks, metric digests, and alert deliveries without blocking primary application threads.
Solutions
- Optimized Elasticsearch document mappings and query DSLs, accelerating retrieval by ~25%.
- Developed a robust C# WinForms application with NSSM background services to handle stream forwarding and dynamic multi-ISP failover.
- Built an automated C# and EmguCV tool featuring barcode-triggered recording (scan-to-start / scan-to-stop) and local rolling disk retention.
- Engineered a decoupled Node.js cron scheduling engine to handle background jobs like email alerts and live Google Sheets data injections.
- Engineered containerized Python edge daemons in Docker interfacing over serial connections with Arduino microcontrollers and camera feeds, paired with a lightweight Next.js operator UI for local access gate overrides.
Impact
- Successfully managed bot configurations and device health telemetry across 200+ enterprise production sites.
- Engineered 45+ customizable dashboards for enterprise analytics visualization.
- Maintained 99% stream uptime despite unpredictable edge network conditions.
- Reduced fraudulent shipping return claims by 90% via automated scan-to-start video verification.
- Seamlessly connected serial Arduino microcontrollers with containerized Python edge daemons for local gate access control.