Evaluation of Large Language Models (LLM) for High-Level Models Generation in Hardware Design H/F

Candidater

The French Atomic Energy and Alternative Energies Commission (CEA) is a major player in research, development and innovation. This technological research organization is active in three main areas: energy, information and health technologies, and defense. Recognized as an expert in its fields, CEA is fully integrated into the European research area and is expanding its presence internationally. The Laboratory for Systems and Technology Integration (LIST), located in the southern Île-de-France region (Saclay), has the mission of contributing to technology transfer and promoting innovation in the field of parallel computing systems.  The Environmental Design and Architecture Laboratory (LECA), within the Digital Systems and Integrated Circuits Department (DSCIN), is a multidisciplinary technological research team comprising experts in hardware IP design and simulation tools. The developed simulation tools rely on various models with different levels of abstraction, tailored to meet the requirements of hardware/software co-design and co-validation. These models are manually crafted through a thorough analysis of specifications (datasheet), making the process tedious and error-prone.  The internship aims to assess the potential of Large Language Models (LLMs) in the hardware design domain, with a specific focus on the challenge of high-level model generation from specifications. It provides an opportunity to explore and leverage the capabilities of LLMs in bridging the gap between modern AI technologies and hardware design.  The intern will undertake a comprehensive evaluation of existing LLMs to determine their suitability for high-level models generation. The first step of the internship is to review and analyze the current state-of-the-art LLMs and their applications in hardware design and code generation. This initial phase is crucial to understand the capabilities and limitations of LLMs, including popular models like GPT-4, and open-source variants. Next, the intern will assess through experiments the suitability of LLMs to identify their potential strengths and weaknesses and their compatibility with high-level model generation. Finally, the project will focus on optimizing the performance of the most promising LLM chosen in the earlier phases. This step involves gathering and preparing relevant datasets and fine-tuning the selected pre-trained LLM to improve its ability to generate accurate high-level models. During this internship, the candidate will gain expertise in evaluating and fine-tuning LLMs for high-level model generation and HW/SW co-design.

The Environmental Design and Architecture Laboratory (LECA), within the Digital Systems and Integrated Circuits Department (DSCIN), is a multidisciplinary technological research team comprising experts in hardware IP design and simulation tools.

Level required:       Master's degree Skills:       Basic understanding of AI, especially Large Language Models (LLMs), and knowledge of HW architecture design would be a plus.       Python, C/C++       English, teamwork, curiosity

Bac+5 - Master of Science

Anglais Intermédiaire

fr_FRFR

Contact us

We will reply as soon as possible...