R&D Engineer at IMT Atlantique

Feedback mechanisms for Edge-to-Cloud applications

During a one-year project at STACK Team - IMT Atlantique, I worked as a Research and Development Engineer on feedback mechanisms for Edge-to-Cloud applications,
under the supervision of Daniel Balouek and Baptiste Jonglez.

This project was carried out in collaboration with INRIA and deployed entirely on the large-scale experimentation platform Grid’5000.
It focused on designing, implementing, and evaluating adaptive systems capable of reacting to runtime changes through feedback loops.

⚙️ Technical Scope

Over the course of the project, I worked with a wide range of tools and technologies, covering the full edge-to-cloud lifecycle:

  • Containerization & Orchestration:

    • Docker
    • k3s / Kubernetes

  • Monitoring & Observability:

    • Prometheus (metrics collection)
    • Grafana (dashboards)
    • OpenTelemetry (instrumentation)
    • Zipkin / Jaeger (distributed tracing)

  • Chaos Engineering & Testing:

    • Chaos Mesh (fault injection and resilience testing)

  • Infrastructure & Deployment:

    • Helm charts
    • CI/CD workflows
    • YAML-based configuration for microservices

💡 Use Case Development

As part of this work, I developed a fully functional use case illustrating feedback mechanisms across edge and cloud environments.
The full project is publicly available on INRIA’s GitLab:
👉 INRIA GitLab – edge-to-cloud-video-processing

This use case includes:

  • A motion detection module at the edge
  • An object recognition module in the cloud using a YOLO model
  • Instrumentation and monitoring of components to track latency, throughput, and resource usage
  • Feedback mechanisms to react to SLO violations (e.g., frame skipping, resource reallocation)

🎯 Outcome

This project allowed me to:

  • Design adaptive edge-to-cloud systems from scratch
  • Integrate modern observability and chaos engineering practices
  • Use Grid’5000 for reproducible experiments at scale
  • Collaborate in an academic research environment with real-world constraints

It was a rich and rewarding experience combining distributed systems, cloud-native engineering, and research-driven software development.