Workshop on Advanced LLM Techniques & AI Architectures
7th - 10th April, 2025
AInnov8 Club, Dronacharya Group of Institutions, Greater Noida organized a4 day workshop on “Advanced LLM Techniques & AI Architectures (A Toolkit for GenAI Architects: Building Smarter LLM Solutions) from 7th - 10th April 2025. The keynote speaker for the workshop was Dr. Souvick Das (Research Associate, Ca’ Foscari University, Venice, Italy). Ms. Kajol Kathuria (Assistant Professor, CSIT) was the coordinator of the workshop.
The main objective of the workshop was to equip students with both foundational and advanced knowledge in transformer models, prompt engineering, retrieval-augmented generation (RAG), semantic search, LangChain, and autonomous AI agents.
Day 1: The Basics, Semantic Search, and Few-shot Learning
The workshop commenced with an introduction to foundational concepts in AI and LLMs. The participants were introduced to the basics of language models and their capabilities. A detailed session on semantic search highlighted how contextual understanding enhances information retrieval. The day concluded with a session on zero-shot and few-shot learning, showcasing the power of LLMs with minimal data examples.
Day 2: Document Processing and Retriever Development
Day 2 focused on practical aspects of document processing. Participants learned how to extract, clean and structure data from diverse document formats. The second half of the day was dedicated to developing retrievers that use semantic embedding techniques for high-performance data retrieval. Real-time exercises with tools such as Lang Chain were a key highlight.
Day 3: Utilizing Generative LLM and RAG Use Cases
The third day explored generative capabilities of LLMs, focusing on generating context-aware and coherent responses. Participants were guided in building end-to-end retrieval-augmented generation (RAG) pipelines. Practical use cases, such as research assistants and intelligent Q&A systems, were discussed.
Day 4: AI Agents and Multi-Agent Frameworks
The final day introduced AI agents capable of reasoning, tool usage, and autonomous retrieval. Participants explored multi-agent frameworks that coordinate multiple AI agents to perform complex tasks. Final project demonstrations included building agents capable of retrieving, processing, and responding intelligently to user queries in academic and industry scenarios.
The 4-day workshop successfully bridged the gap between theoretical concepts and real-world AI applications. Participants gained not only conceptual clarity but also practical exposure to building LLM-powered applications. The workshop concluded with enthusiastic feedback and a vision for continuing education in AI innovations.