Internship for Retriever LLM Evaluation Test on Electrical Drive Test Data

Institute

Lehrstuhl für Nachhaltige Mobile Antriebssysteme (TUM-ED)

Type

Bachelor's Thesis / Semester Thesis /

Content

experimental /  

Description

In a LLM-powered retrieval-augmented generation (RAG) system, retriever models impact the final answer generation quality of LLM. As a part of post-training of search agent for electrical drive test document, the performance of retriever models as well as the functionability of reranker models need to be investigated in a standard RAG framework. This internship focuses on experiments in comparing and evaluating the performance limits of retriever models across various factors: parameter, size, combination with re-ranking models, etc. on both public datasets and private data. The project can only be provided as internship to be conducted full-time in 9 weeks according to FSPO of MSEI, BSEI and other equivalent degree programs.

 

Your task:

• Cleansing of open-source datasets and private data
• Configuration of RAG system with various retriever models
• Building experiment platform and perform experiments
• Collecting results and analysis

Requirements

• Interest in Artificial Intelligence, Large Language Models, and Agent-based Systems
• Basic Python programming skills
• Experience in LLM projects is welcome
• MSEI, BSEI and other equivalent degree programs

Possible start

sofort

Contact

M.Sc. Kai Cui
Room: 2107.EG.008
Phone: +49 8928924108
k.cuitum.de