AI Computer Vision Engineer

About Us We are building a next-generation AI Retail Intelligence Platform that transforms existing retail CCTV cameras into intelligent business assistants. Retail stores already have CCTV cameras installed for security. Our goal is to use those cameras to generate real-time business intelligence without requiring any new camera hardware. This is a long-term startup project. We are looking for a highly experienced AI Computer Vision Engineer to build the first production-ready version of our platform. If the collaboration is successful, you will become a long-term technical partner and Founding AI Engineer. Project Vision Our platform should connect directly to the retailer's existing CCTV system and analyze video streams in real time. Instead of using CCTV only for recording, our AI will understand customer movement, detect shopping behavior, generate events, identify repeat customers (when customer enrollment and consent exist), and provide actionable business insights. This is NOT a security surveillance project. This is NOT an attendance system. This is NOT just a face recognition project. We are building an AI-powered Retail Intelligence Platform. Phase 1 – Video Pipeline & Computer Vision Foundation The system must support: Existing CCTV cameras RTSP Streams ONVIF Protocol Hikvision Dahua CP Plus Axis Uniview Generic IP Cameras USB Cameras (for testing) The AI engine should: Connect automatically Reconnect after network failures Support multiple camera streams Process video in real time Maintain low latency Be production ready Support Docker deployment Work on Linux and Windows Be designed for future Edge AI deployment (NVIDIA Jetson preferred) Phase 1 – AI Detection Implement: Person Detection Face Detection Multi-Person Detection Face Tracking Multi-Object Tracking Person Tracking Entry Detection Exit Detection Occlusion Handling Detection Confidence Score Preferred technologies: Python OpenCV PyTorch YOLO ONNX Runtime TensorRT ByteTrack / DeepSORT / StrongSORT (or equivalent) Phase 2 – Event Engine The system should convert AI detections into meaningful business events. Examples: Customer Entered Customer Left New Visitor Returning Visitor Customer Waiting Long Waiting Customer Queue Started Queue Ended Customer Stayed More Than X Minutes Camera Offline Low Confidence Detection Every event should include: Timestamp Camera ID Tracking ID Confidence Score Event Type Expose these events through REST APIs or WebSockets. Phase 2 – Analytics Engine Generate retail analytics including: Total Visitors Returning Visitors New Visitors Peak Hours Customer Flow Average Visit Duration Store Traffic Timeline Customer Frequency Daily / Weekly / Monthly Analytics The analytics engine should be modular and scalable. Phase 3 – Customer Intelligence This is the core feature of our platform. The AI should support customer re-identification only when customer enrollment and consent are available. The system should integrate with our CRM/POS through APIs to retrieve customer information. Example customer profile: Customer Name Customer ID Total Visits Last Visit Average Visit Duration Lifetime Purchase Average Bill Value Favourite Brand Favourite Category Preferred Shopping Time Birthday Membership Status Example dashboard event: 🔔 Customer Arrived Returning Customer Visits: 23 Lifetime Purchase: ₹4,75,000 Average Bill: ₹21,500 Last Visit: 18 Days Ago Favourite Brand: Nike Birthday: Next Week The AI engine should never store payment information directly. It should retrieve customer information from the retailer's CRM/POS using secure APIs. Dashboard Integration The AI engine should expose APIs for frontend integration. The dashboard should display: Live Customer Events Customer Profiles Visitor Analytics Heatmaps (future-ready) Queue Statistics Customer Timeline Camera Status Real-Time Alerts Performance Requirements The solution must: Run in real time Low latency GPU optimized Modular architecture Scalable codebase Clean documentation Docker support Production-ready deployment Future Roadmap (Architecture Ready) Although not required in this milestone, the architecture should support future features such as: Shelf Analytics Product Interaction Detection Queue Intelligence Theft Detection Staff Analytics Multi-Store Management AI Business Recommendations WhatsApp Alerts Mobile App Integration Edge AI Deployment Required Skills Python Computer Vision OpenCV PyTorch YOLO ONNX TensorRT Multi-Object Tracking Face Detection RTSP ONVIF REST APIs Docker Linux Git Preferred: NVIDIA Jetson Hailo AI OpenVINO Retail Analytics Experience Questions for Applicants Have you built production-grade Computer Vision systems? Have you worked with RTSP and CCTV integrations? Which object tracking framework would you recommend and why? Have you built customer analytics or retail analytics solutions? Have you deployed AI on NVIDIA Jetson or other edge devices? Share links to similar projects or GitHub repositories. How would you architect this platform to support thousands of retail stores? What challenges do you foresee, and how would you solve them? Long-Term Opportunity This is not a one-time freelance project. We are building a global AI Retail Intelligence platform and are looking for a long-term technical partner who can help us scale the product into a world-class AI SaaS platform. "The ideal solution should be privacy-aware by design. The architecture should support configurable identity features, allowing retailers to use anonymous analytics only, or customer-linked analytics where customers have explicitly enrolled and consented.