Project Overview:

This project explores how state-of-the-art AI models, such as Large Language Models (LLMs), can simplify the process of running circuit simulations. Typically, creating simulation files for tools like NGSPICE [1] requires significant manual effort and technical expertise. The goal is to develop a local assistant, powered by Meta's Llama 3-8B [2], that can automatically generate these files using effective prompt engineering

The student will learn how to work with LLMs locally, apply basic prompt engineering, and use retrieval-augmented generation (RAG) [3] to improve prompts with less manual effort. Circuit simulation will serve as the main application example, but no deep knowledge of electronics is required.

Through this project, students will gain practical experience in:
      • How to use LLMs for real-world automation tasks
      • How to apply prompt engineering and retrieval techniques
      • Basics of circuit simulation (with guidance)

Minimum Requirements:

      • Solid knowledge of Python and Linux

Preferred Additional Skills:

      • Basic understanding of LLMs and electronic circuit simulation

References:

[1] Vogt, Holger, et al. "Ngspice User’s Manual Version 34 (ngspice release version)." 31 Jan. 2021
[2] Grattafiori, Aaron, et al. "The llama 3 herd of models." arXiv preprint arXiv:2407.21783 (2024).
[3] Lewis, Patrick, et al. "Retrieval-augmented generation for knowledge-intensive nlp tasks." Advances in neural information processing systems 33 (2020): 9459-9474.