RAG_based_Chatbots_AngelHack.pdf

Transcript

1. Abhiram Ravikumar | @abhi12ravi Blank for Cover / Divider Building RAG based chatbots May 2024 Angel Hack Series
2. Abhiram Ravikumar | @abhi12ravi Defining terms 2 RAG LLM Vector Databases Langchain
3. Abhiram Ravikumar | @abhi12ravi What is RAG? 3
4. Abhiram Ravikumar | @abhi12ravi RAG architecture 4
5. Abhiram Ravikumar | @abhi12ravi RAG components 5 - Retriever - Ranker - Generator - External Data
6. Abhiram Ravikumar | @abhi12ravi RAG components 6 External Data - Data outside LLM’s training data - Vector embeddings - Text to vector form - Cohere model (from Bedrock)
7. Abhiram Ravikumar | @abhi12ravi RAG components 7 Vector DB - Stores embedding - Data chunking is applied - Open source - Chroma, Qdrant, many others
8. Abhiram Ravikumar | @abhi12ravi RAG components 8 RAG Retriever - Relevancy search - Identification and extraction of required text from a large corpus
9. Abhiram Ravikumar | @abhi12ravi RAG components 9 RAG Ranker - Refines retrieved info based on: - Relevance - Importance
10. Abhiram Ravikumar | @abhi12ravi RAG components 10 RAG Generator - LLM (GPT, Claude, etc)
11. Abhiram Ravikumar | @abhi12ravi RAG based Chat application 11
12. Abhiram Ravikumar | @abhi12ravi RAG vs Fine Tuning 12 Training data • Fine-tuning requires task-specific labelled data examples • More time and cost • RAG relies on pre-trained LLM & external knowledge bases Adaptability • LLM remains generalized in the case of RAG, whereas fine-tuning makes LLM more specialized and tailored to specific tasks Model Architecture • In RAG, LLM architecture remains the same but in fine-tuning, params of pre- trained LLMs are modified
13. Abhiram Ravikumar | @abhi12ravi Potential project ideas 13 Digital Empowerment • FAQ Chatbot for BMTC / Metro routes • Volunteer matching chatbot for NGOs • Study buddy chatbot • Mental wellbeing chatbot Financial Inclusion • Financial literacy chatbots for rural and youth • Savings Goals tracker chatbot • Credit Score Advice chatbot Social Responsibility • Recycling Guidelines Chatbot • Green initiatives participation chatbot • Environment impact analyzer bot
14. Abhiram Ravikumar | @abhi12ravi References 14 - Intro to RAG apps - https://www.slideshare.net/slideshow/introduction-to-rag-retrieval- augmented-generation-and-its-application/266746505 • Confluence Bot docs - https://github.com/BastinFlorian/RAG- Chatbot-with-Confluence
15. Abhiram Ravikumar | @abhi12ravi Blank for Cover / Divider Q & A Abhiram Ravikumar Email: abhi12ravi@gmail.com LinkedIn: in.LinkedIn.com/in/abhi12ravi/