Backend Engineer – Real-Time Video Processing & n8n Automation

We are looking for an experienced Backend Engineer to build a complete video analytics pipeline capable of processing video streams (from files), performing GPU-accelerated AI inference, tracking objects over time, and orchestrating intelligent alerts through n8n. The main objective is to develop a reliable system that ingests video data, performs detection and tracking, and generates Telegram notifications while preventing duplicate alerts and managing application state. This is a POC project. Testing will be done using pre-recorded video files, not live RTSP streams. The architecture must be capable of handling real-time streaming (RTSP-ready), but validation will use video files. The technology stack includes Python, OpenCV, FFmpeg, YOLOv8n, ByteTrack, PostgreSQL, n8n, Docker, and RunPod. Duration: maximum 3 to 4 weeks Scope of Work Infrastructure and Environment Setup - Deploy and configure the development environment on GPU infrastructure using RunPod - Set up a managed PostgreSQL database such as Supabase or an equivalent solution - Deploy and configure n8n in either a cloud or self-hosted environment - Create a production-ready Docker container for the Python backend service - Implement centralized logging and monitoring for system events Video Processing Pipeline - Develop a pipeline capable of ingesting and processing video files using OpenCV and FFmpeg - The pipeline architecture must support RTSP streaming for future deployment, but will be tested and validated using pre-recorded video files provided for testing - Integrate YOLOv8n for person and object detection with confidence scoring - Generate bounding boxes and structured detection outputs - Benchmark the solution on one to two hours of test video data to evaluate latency and detection accuracy - Produce standardized JSON payloads that will be transmitted to n8n through webhooks Object Tracking and State Management - Integrate ByteTrack to maintain consistent object identities across frames - Implement configurable Regions of Interest (ROI) using polygon-based zones - Maintain object state information, including movement and stationary status - Build a detection-to-webhook pipeline that sends every relevant event to n8n n8n Workflow Development - Design robust n8n workflows responsible for processing incoming detection events - Implement webhook ingestion and JSON parsing - Create deduplication logic to suppress repeated alerts within a configurable time window - Implement rate limiting to comply with Telegram API limitations - Manage alert history and state persistence through PostgreSQL - Trigger Telegram notifications with formatted messages and associated snapshot images - Support multiple notification channels based on configurable monitoring zones - Implement resilient error handling: - Automatic retries for failed webhook executions - Logging of unsuccessful notifications - Graceful handling of Telegram API errors Testing and Validation - Validate end-to-end functionality from object detection through Telegram notification delivery using provided video files - Perform load testing with more than 100 webhook events per second - Ensure notification latency remains below two seconds under normal processing conditions - Produce complete technical documentation and deployment instructions Deliverables - Complete source code hosted in a GitHub repository - Production-ready Docker image - Exportable and reusable n8n workflow JSON files - Deployment guide covering RunPod, PostgreSQL, and n8n installation - Performance benchmark report based on one to two hours of video file testing - Telegram bot configuration guide and setup instructions - Troubleshooting documentation covering common deployment and workflow issues Required Skills - Strong experience with Python, including asynchronous programming, multithreading, and robust error handling - Hands-on expertise with n8n workflow development, including webhooks, conditional logic, retries, and exception management - Solid knowledge of PostgreSQL schema design, transactions, and state management - Experience building and optimizing Docker containers for production environments - Comfortable working in Linux environments with command-line tools and system debugging - Experience integrating the Telegram Bot API for notifications and image delivery - Practical knowledge of OpenCV, FFmpeg, and deep learning inference pipelines using YOLO - Experience with object tracking algorithms such as ByteTrack is highly preferred Ideal Candidate The ideal candidate has experience building production-grade computer vision systems and workflow automation platforms. They are comfortable working independently, delivering clean and maintainable code, and producing reliable solutions that can process continuous video streams efficiently. For this project, testing will be conducted using pre-recorded video files. The candidate should be able to validate the system works correctly with these test files and is architected to support real-time RTSP streams in the future.