Edge Computing System Characterization

As edge computing becomes a cornerstone of modern IoT and AI applications, understanding its operational characteristics is critical for designing efficient, reliable, and scalable systems. This internship focuses on the systematic characterization of edge computing platforms, evaluating key parameters such as latency, throughput, energy consumption, resource utilization, and security overhead. The project will involve benchmarking diverse hardware and software configurations under real-world workloads, analyzing trade-offs between performance and power efficiency, and identifying optimization opportunities for edge-based AI inference and data processing. By the end of the internship, participants will deliver a comprehensive performance profile and actionable insights that contribute to advancing edge computing technologies for next-generation applications.

The internship will focus on evaluating and characterizing edge computing systems in terms of performance, energy efficiency, and reliability for AI and IoT workloads. The scope includes:

• Hardware Platforms: ARM-based SoCs, embedded GPUs, and FPGA accelerators.
• Software Frameworks: TensorFlow Lite, ONNX Runtime, and custom edge inference engines.
• Workload Types: AI inference (vision, NLP), real-time IoT data processing.
• Benchmarking Dimensions:
  • Latency and throughput under varying loads.
  • Power consumption and energy per inference.
  • Resource utilization (CPU/GPU, memory footprint).
  • Network performance (bandwidth, jitter).
  • Accuracy vs. performance trade-offs.
• Security & Reliability: Impact of encryption and fault recovery on system performance.

By the end of the internship, the student will:

• Understand edge computing architectures and their role in IoT and AI ecosystems.
• Gain hands-on experience in benchmarking edge systems for performance, energy efficiency, and reliability.
• Learn to use profiling tools and frameworks (e.g., MLPerf Edge, TensorFlow Lite, iperf) for system characterization.
• Develop skills in data analysis and visualization for interpreting benchmarking results.
• Explore trade-offs between latency, power, and accuracy in edge deployments.
• Acquire knowledge of security and scalability considerations in edge computing environments.

Required Skills

• Programming: Proficiency in Python and C/C++ (for embedded or system-level tasks).
• Data Analysis: Ability to process and visualize performance metrics using tools like Pandas, Matplotlib.
• Basic Hardware Knowledge: Familiarity with ARM-based platforms, GPUs, or single-board computers (e.g., Raspberry Pi, Jetson platforms).
• Networking Fundamentals: Understanding of bandwidth, latency, and communication protocols.
• Machine Learning Basics: Experience with lightweight ML frameworks (TensorFlow Lite, ONNX Runtime) is a plus.
• Problem-Solving & Documentation: Ability to design experiments and clearly report findings.

Type of Internship: Master internship; Combination of internship and thesis; Internship; Thesis

Master's degree: Master of Engineering Technology; Master of Engineering Science

Required educational background: Computer Science

Duration: 1 year

University Promotor: Marian Verhelst (KU Leuven)

For more information or application, please contact the supervising scientist Cagatay Ozdemir (Cagatay.Ozdemir@imec.be).

Imec allowance will be provided for students studying at a non-Belgian university.