Feedback Mechanisms for Edge-to-Cloud Applications at IMT Atlantique

On December 1st, 2023, I joined IMT Atlantique as a Research and Development Engineer to work on feedback mechanisms for Edge-to-Cloud video processing applications.

The project focused on improving the monitoring, analysis, and adaptation of distributed applications across heterogeneous environments. I had the opportunity to design and implement a full feedback loop that connects metrics collection, performance monitoring, and automated reactions.

🔍 Project Objectives

• Improve observability of distributed video pipelines (motion detection, object recognition)
• Enable dynamic feedback to ensure performance and quality objectives are met
• Detect and respond to SLO violations through actionable insights

⚙️ My Contributions

• Designed a monitoring pipeline using Prometheus and OpenTelemetry
• Implemented metrics collection on FPS, latency, processing time, and object recognition quality
• Developed a Flask-based interface for visualizing critical metrics and system health
• Created a prototype integrating:
  • A camera simulator
  • Motion detection at the edge
  • Cloud-based object recognition (YOLOv5)
• Traced processing flow end-to-end using distributed tracing

🛠️ Tools & Technologies

• Python, Flask
• Prometheus & OpenTelemetry
• Docker, Docker Compose
• GitLab CI/CD
• YOLOv5, OpenCV
• Grafana (for metric dashboards)

📘 Outcomes

• Developed and documented an open-source prototype:
  GitLab Repository
• Explored potential extensions using critical metric detection and SLI/SLO reasoning
• Shared findings internally and supported future directions at IMT Atlantique and Inria

---

This experience helped shape my current work on resilient, adaptive applications, and provided a solid foundation for my ongoing research on intelligent configuration and monitoring systems.

Feel free to reach out if you’re working on similar topics!