| Hora | Sesión / Ponente / Auditorio |
|---|---|
| 09:00 |
Registro
Auditorio Salvador Allende
|
| 09:30 |
Bienvenida
Auditorio Salvador Allende
|
| 10:00 10:30 |
Auditorio Salvador Allende
Talk
On Private: Orchestrating Gemma 4 and AppFunctions on Android Ver descripción de la charlaHere is Dinorah, your trustable Android Developer that also loves security As Android evolves into an AI-first OS, the role of a health app is shifting from a passive data silo to an active, intelligent agent. But how do we build these "agentic" experiences without compromising the most sensitive data a user owns? This session explores the technical intersection of the new AppFunctions API and Gemma 4 (E2B/E4B) models. We will dive into architecting a system where on-device LLMs reason over Health Connect data locally via AICore, ensuring zero-cloud roundtrips. Keep you health private and secure |
| 11:00 11:30 |
Auditorio Salvador Allende
Workshop
Offensive cloud architecture: Building adversarial infraestructure on GCP |
| 12:00 12:30 |
Auditorio Salvador Allende
Workshop
Pedro Lara
Fine-Tuning Gemma for Fact Retrieval: A MeMo-Based Approach to Next-Gen RAG Systems Ver descripción del tallerThis presentation explores a practical, end-to-end implementation of the Memory Model (MeMo) architecture, offering a compelling approach for developers building advanced agentic pipelines or enhancing Large Language Model (LLM) memory beyond traditional Retrieval-Augmented Generation (RAG). Attendees will be guided through a comprehensive Colab notebook demonstrating a dual-model framework that separates reasoning from factual recall. The session covers the complete technical pipeline—from automated fact extraction, semantic consolidation, and cross-document synthesis for multi-hop reasoning—to the final execution of a multi-turn inference loop orchestrating Gemini as the Executive model for question decomposition and synthesis, alongside a LoRA-fine-tuned Gemma via KerasHub acting as the parameterized Memory Model.
Auditorio Adalberto Navarro
Workshop
Vibe Code with Gemini in Google AI Studio Ver descripción del taller¡Vamos a aprender programando un juego! Crearemos juntos "Snake & Beats", un juego retro con estética neón y reproductor de música integrado. Utilizaremos Gemini en Google AI Studio para generar una aplicación de React completa en un solo archivo y te enseñaré cómo realizar el despliegue directo en Cloud Run para que tu proyecto esté disponible en la web. |
| 13:00 13:30 |
Auditorio Salvador Allende
Talk
Diferencia entre vibecoding e ingenieria asistida por IA |
| 14:00 |
|
| 14:30 |
Cierre
Auditorio Salvador Allende
|
| 15:00 |
Networking
Auditorio Salvador Allende
|