Agentic RAG Planner
Develop a planner that is able to giude the Retrieval Augmented Generation (RAG) process, into answering complex queries.
Requirements
- M.Sc. in Machine Learning, Data Science, Computer Science, Mathematics, Telecommunications, or similar
- Knowledge of Python
- Software development skills
- Knowledge of natural language understanding
- Basic knowledge of retrieval, semantic embedding, RAG
Description
Naive RAG answers queries by retrieving the top-K semantically related paragraphs from a large document collection. This approach works well for some types of factual queries, but it is less effective for more complex ones, such as general, comparative, or cross-document questions, e.g.: ‘how many times is A mentioned in the documents?’, ‘pros and cons of electric cars’, ‘what do all these documents talk about?’. The goal of this internship is to explore a planning-based RAG agent that can decompose an original query into a sequence of intermediate steps, enabling it to answer complex queries more effectively.