The talk focuses on the Autoregressive model architecture and why there is a need for the decision model in this current scenario of the agentic era and what DiffusionGemma provides, so the aim of mine is to introduce and talk about the experimentation of the decision model as DiffusionGemma and Jev and what one can build with this approach.

Also will have the discussion around how one can implement it well using the Google Cloud.

Agenda for the talk

- Understanding of how LLM works (token generation)
- Why need of the "System one" model, why now ?
- Moving from generating text to making decisions
- LLM as Judge, Useful ? but how reliable ?
- Understanding Jev Approach
- Exploring DiffusionGemma (djev)
- Understanding the Jev inDepth
- Experimentation with the DiffusionGemma and Jev
- Best Approach to make use via Google Cloud
- What future looks like
- Q&A

No Pre-requisite needed for the talk, It takes from beginner to intermediate level