GTech Hybrid-Cloud Automation Infrastructure

Secure SSH Bridge for High-Throughput VPS-to-Cloud Image Processing

A production-grade hybrid-cloud automation system that seamlessly bridges VPS (Virtual Private Server) environments with RunPod cloud compute infrastructure. Engineered a custom SSH bridge with rsync protocol to enable secure, high-performance image processing workflows between on-premise storage and cloud computation units.

50,000+
Images Processed
75%
Processing Time Reduction
99.5%
System Uptime

Project Overview

This system is a production-ready hybrid-cloud automation infrastructure designed to securely connect client VPS environments with GPU-accelerated cloud compute resources. Successfully deployed for GTech's enterprise image processing workflows, it demonstrates advanced capabilities in secure file transfer, intelligent deduplication, and automated processing orchestration.

Key Achievements

  • ✅ Processed 50,000+ images with zero security incidents
  • ✅ Reduced processing time by 75% vs. local infrastructure
  • ✅ Achieved 99.5% system uptime with automatic retry mechanisms
  • ✅ Implemented permanent file deduplication (zero re-processing)
  • ✅ SSH key authentication with ControlMaster connection pooling

Primary Use Cases

  • 🔄 Automated VPS-to-Cloud file synchronization via rsync
  • 🖼️ High-throughput image processing with GPU acceleration
  • 🔐 Secure SSH bridge with encrypted file transfers
  • 📊 Intelligent history tracking and duplicate prevention
  • ☁️ Bidirectional transfers: fetch → process → send back
  • 🎯 Path normalization for cross-system compatibility

Key Features

Enterprise Security

SSH key-based authentication with ControlMaster for persistent connections. Supports password fallback with sshpass. All transfers encrypted over SSH tunnels with no plaintext credentials in logs.

Bidirectional File Sync

rsync-based transfers with --protect-args for space handling, --files-from for selective batch operations, and compression for bandwidth optimization. Handles 10-20 MB/s throughput.

Permanent Deduplication

History tracking with permanent ban lists prevents re-fetching/re-sending processed files. Cross-folder matching maps input to output directories. Automatic log rotation at 10MB threshold.

Smart Path Normalization

Automatically handles folder name variations: spaces→underscores, IP_→OP_ prefix replacement, date format conversion (YY→2026), and special character sanitization for cross-system compatibility.

GPU-Accelerated Processing

RunPod integration with NVIDIA A40/A100 GPUs for AI-powered background removal. Supports Rembg (U2-Net), BIREFNET (Transformer), and Recraft API. Processes 50-100 images/hour with batch parallelization.

Automated Retry & Recovery

Connection state caching for reuse, exponential backoff for retries, timeout handling for failed operations, and comprehensive logging for troubleshooting. 95% retry success rate after initial failures.

Technology Stack

Infrastructure & DevOps

SSH Protocol rsync ControlMaster sshpass Linux Ubuntu RunPod GPU Instances

Backend & Processing

Python subprocess threading Gradio Pillow (PIL) OpenCV

AI/ML Models

Rembg (U2-Net) BIREFNET Recraft API EasyOCR Transformers PyTorch

Results & Business Impact

Quantified Results

  • 18,000+ images processed in 1 month across GTech projects
  • 75% faster processing compared to local infrastructure
  • 99.5% system uptime with automatic error recovery
  • Zero security incidents with SSH key authentication
  • 10-20 MB/s transfer speeds with rsync compression
  • 98.5% success rate for image processing operations

Technical Highlights

  • SSH ControlMaster for connection pooling and reuse
  • Permanent ban lists prevent duplicate processing (zero waste)
  • Path normalization handles cross-system naming variations
  • RAW file support (CR2, NEF, ARW) with auto-conversion
  • Gradio web interface for real-time monitoring
  • Automatic log rotation at 10MB threshold

Interested in Hybrid-Cloud Solutions?

Let's discuss how secure cloud infrastructure can transform your processing workflows

← Back to Portfolio Contact Me →