6:45 PM: Harshit Meta - nsurance Policy Chatbot with Google Cloud Gemini and MongoDB Atlas Vector Search
In this talk, we will demo an insurance policy chatbot. The chatbot will use Retrieval Augmented Generation (RAG) to ground LLM responses. The tech stack includes Gemini (LLM), Google Cloud’s embedding model (vector generation), and MongoDB Atlas (vector search).
7:15 PM: Tony O'Halloran - Cad é an Ghaeilge ar 'Transformer'? An Introduction to the Transformer Architecture for Translation
In this session, we’ll use Python to build our own model capable of (really poorly) translating from English to Irish. We’ll start with the basics of deep learning and build up an understanding of how the transformer architecture works for sequence-to-sequence tasks.
7:45 PM: Urja Pawar - Exploring AI Decisions: A Deep Dive into Google’s What-If Tool
Understanding and trusting AI models is crucial for their effective deployment. Google’s What-If Tool offers an interactive and intuitive way to visualize and analyse machine learning models’ behaviour. This session will provide a comprehensive overview of the What-If Tool, demonstrating how it can be used to inspect model performance, explore counterfactuals, and ensure fairness across different datasets. Participants will learn to utilise this tool for debugging, auditing, and improving their models, making AI systems more transparent and reliable.
