AI Smart Shoplifting System QA Review
Budget: $30 – $250 USD
PROJECT SPECIFICATION: Technical Audit & QA Verification
Role: Independent Technical Auditor / QA Engineer
Project: AI-Based Smart Shoplifting Detection System
Objective: To verify the quality, security, and performance claims made by the Lead Developer before milestone payments are released.
1. Role Overview
You will act as the Client’s technical representative. Your job is not to write code, but to audit, test, and verify the deliverables submitted by the Lead Developer. You must confirm that the system meets the strict requirements for Accuracy (>85%), Latency (<10s), and Security (Encryption).
2. Responsibilities by Milestone
You will perform a "Pass/Fail" inspection at the end of each phase. The Lead Developer will not receive payment until you sign off.
Phase 1 Audit: Architecture & Connectivity
• Verify Connection: Log in to the Jetson device remotely and confirm it is successfully pulling RTSP streams from the NVR.
• Code Review (Architecture): Review the proposed system design to ensure it is scalable (supports the "Brand -> Branch" tree structure) and not hard-coded for a single store.
• Deliverable: Phase 1 Inspection Report (Pass/Fail).
Phase 2 Audit: MVP Pipeline Verification
• Pipeline Test: Trigger a manual event (walk in front of the camera). Verify that the Mobile App receives a notification.
• Video Check: Confirm the video plays correctly on both iOS and Android test devices.
• Latency Benchmark (Initial): Measure the time from "Event Trigger" to "App Notification." Report the current speed (even if not yet <10s).
Phase 3 Audit: Security & Advanced Features (CRITICAL)
• Latency Stress Test: Use network monitoring tools (e.g., Wireshark or server logs) to strictly measure the "10-Second Rule."
o Requirement: The alert must arrive in <10s.
• Security Audit (Anti-Theft):
o Attempt to Read Code: SSH into the device and try to open the source files. Verify that no raw .py files exist and that files are compiled binaries (.so).
o Encryption Check: Verify that LUKS (Disk Encryption) is active and the storage cannot be mounted on a different Linux machine without the key.
o Hardware Lock Test: Ask the developer to demonstrate the "Hardware ID Check" (e.g., by trying to run the binary on a different dummy Jetson board to ensure it fails).
• Kill Switch Test: Send the "Sleep" command from the Dashboard and verify via system logs (top/htop) that the GPU processes stop instantly.
Phase 4 Audit: Accuracy & Final Stability
• Accuracy Validation:
o Review the logs of the last 100 detected events.
o Compare the "System Detection" vs. "Manager Feedback."
o Calculation: Confirm the mathematical accuracy is >85%.
• Local Learning Verification:
o Test: Mark a specific action as "False Alarm" in Branch A. Ensure that Branch B's model does not change (verifying the "Local Learning" rule).
• Stress Test: Monitor the system uptime logs to confirm it ran for 48 hours without a restart or crash.
3. Required Access & Tools
To perform this job, the Client will provide you with:
1. Read-Only access to the Source Code Repository (GitHub/GitLab).
2. SSH Access to the testing Jetson device.
3. Access to the Web Dashboard as an Admin.
4. Documentation provided by the Lead Developer.
4. Deliverables
For every phase, you must submit a "Milestone Verification Report" containing:
1. Test Results: (e.g., "Average Latency: 8.4 seconds - PASS")
2. Security Vulnerabilities: (e.g., "Found raw config file with passwords - FAIL")
3. Final Recommendation: "Approve Payment" or "Reject & Require Fixes."
Main project description are attached .
Role: Independent Technical Auditor / QA Engineer
Project: AI-Based Smart Shoplifting Detection System
Objective: To verify the quality, security, and performance claims made by the Lead Developer before milestone payments are released.
1. Role Overview
You will act as the Client’s technical representative. Your job is not to write code, but to audit, test, and verify the deliverables submitted by the Lead Developer. You must confirm that the system meets the strict requirements for Accuracy (>85%), Latency (<10s), and Security (Encryption).
2. Responsibilities by Milestone
You will perform a "Pass/Fail" inspection at the end of each phase. The Lead Developer will not receive payment until you sign off.
Phase 1 Audit: Architecture & Connectivity
• Verify Connection: Log in to the Jetson device remotely and confirm it is successfully pulling RTSP streams from the NVR.
• Code Review (Architecture): Review the proposed system design to ensure it is scalable (supports the "Brand -> Branch" tree structure) and not hard-coded for a single store.
• Deliverable: Phase 1 Inspection Report (Pass/Fail).
Phase 2 Audit: MVP Pipeline Verification
• Pipeline Test: Trigger a manual event (walk in front of the camera). Verify that the Mobile App receives a notification.
• Video Check: Confirm the video plays correctly on both iOS and Android test devices.
• Latency Benchmark (Initial): Measure the time from "Event Trigger" to "App Notification." Report the current speed (even if not yet <10s).
Phase 3 Audit: Security & Advanced Features (CRITICAL)
• Latency Stress Test: Use network monitoring tools (e.g., Wireshark or server logs) to strictly measure the "10-Second Rule."
o Requirement: The alert must arrive in <10s.
• Security Audit (Anti-Theft):
o Attempt to Read Code: SSH into the device and try to open the source files. Verify that no raw .py files exist and that files are compiled binaries (.so).
o Encryption Check: Verify that LUKS (Disk Encryption) is active and the storage cannot be mounted on a different Linux machine without the key.
o Hardware Lock Test: Ask the developer to demonstrate the "Hardware ID Check" (e.g., by trying to run the binary on a different dummy Jetson board to ensure it fails).
• Kill Switch Test: Send the "Sleep" command from the Dashboard and verify via system logs (top/htop) that the GPU processes stop instantly.
Phase 4 Audit: Accuracy & Final Stability
• Accuracy Validation:
o Review the logs of the last 100 detected events.
o Compare the "System Detection" vs. "Manager Feedback."
o Calculation: Confirm the mathematical accuracy is >85%.
• Local Learning Verification:
o Test: Mark a specific action as "False Alarm" in Branch A. Ensure that Branch B's model does not change (verifying the "Local Learning" rule).
• Stress Test: Monitor the system uptime logs to confirm it ran for 48 hours without a restart or crash.
3. Required Access & Tools
To perform this job, the Client will provide you with:
1. Read-Only access to the Source Code Repository (GitHub/GitLab).
2. SSH Access to the testing Jetson device.
3. Access to the Web Dashboard as an Admin.
4. Documentation provided by the Lead Developer.
4. Deliverables
For every phase, you must submit a "Milestone Verification Report" containing:
1. Test Results: (e.g., "Average Latency: 8.4 seconds - PASS")
2. Security Vulnerabilities: (e.g., "Found raw config file with passwords - FAIL")
3. Final Recommendation: "Approve Payment" or "Reject & Require Fixes."
Main project description are attached .