The growing integration of AI services requires further improvements in the energy efficiency and performance of AI accelerators. The AI hardware simulation ecosystem plays a vital role in designing the next generation of AI accelerators.
Designing and implementing simulators typically requires significant investment of time and resources, which are often unavailable. In the era of AI, this process can potentially be automated by leveraging AI agents.
The goal of this project is to explore whether a set of AI agents can autonomously generate simulators for AI accelerators and iteratively evolve their design. This evolution would be guided by continuous regression testing and correlation with real hardware platforms. The project aims to investigate the feasibility of an AI-driven workflow that not only creates accurate simulation models but also improves them over time based on performance and architectural fidelity.
Key research questions include:
As part of this internship, you will:
Ideal candidate profile:
Master's degree: Master of Engineering Technology, Master of Science, Master of Engineering Science
Required educational background: Computer Science, Electrotechnics/Electrical Engineering
Duration: 12 months (NNFC internship only)
For more information or application, please contact the supervising scientists Tommaso Marinelli (< email deleted for security reasons >), Lev Mukhanov (< email deleted for security reasons >) and Konstantinos Tovletoglou (< email deleted for security reasons >).
Designing and implementing simulators typically requires significant investment of time and resources, which are often unavailable. In the era of AI, this process can potentially be automated by leveraging AI agents.
The goal of this project is to explore whether a set of AI agents can autonomously generate simulators for AI accelerators and iteratively evolve their design. This evolution would be guided by continuous regression testing and correlation with real hardware platforms. The project aims to investigate the feasibility of an AI-driven workflow that not only creates accurate simulation models but also improves them over time based on performance and architectural fidelity.
Key research questions include:
- Can AI agents learn the structure and behavior of AI system-level architectural accelerators well enough to generate functional simulators?
- How effectively can these agents adapt and refine the simulator design using feedback from regression tests and real-world measurements?
- What methodologies and frameworks are required to ensure scalability and reliability of this automated approach?
As part of this internship, you will:
- Define and build a framework of AI agents to generate a simulator for the target AI accelerator (the assistance from AI software engineer will be provided).
- Define and implement regression test framework to iteratively evolve the simulator.
- Investigate inaccuracies in the generated simulators at both the architecture and memory subsystem levels.
- Summarize the identified inaccuracies across various AI workloads.
- Propose techniques to improve the AI-based generation framework.
Ideal candidate profile:
- MSc student in Computer Science, Electrical Engineering, or a related program.
- Strong understanding of computer hardware architecture, memory systems, and AI accelerators.
- Familiarity with neural network architectures, especially LLMs, and awareness of emerging paradigms such as multi-agent systems and diffusion-based models.
- Proficiency in C/C++ and Python programming languages.
- Experience with performance analysis tools or simulation frameworks is a plus
- Self-starter with the ability to work independently and think critically.
- Available for a 1-year internship and eligible to work in Belgium.
- Strong written and verbal English communication skills.
Master's degree: Master of Engineering Technology, Master of Science, Master of Engineering Science
Required educational background: Computer Science, Electrotechnics/Electrical Engineering
Duration: 12 months (NNFC internship only)
For more information or application, please contact the supervising scientists Tommaso Marinelli (< email deleted for security reasons >), Lev Mukhanov (< email deleted for security reasons >) and Konstantinos Tovletoglou (< email deleted for security reasons >).
