DevFest 2024

The most secure server is one that is disconnected from the Internet and unplugged. And the most convenient environment for devs is where they have admin access to production and the freedom do what they want. How do you marry the two? Let's look at a real-world scenario where we built a cloud-native fintech platform on GKE. The vision? A robust, flexible, and secure foundation that supports SOC2-compliant deployments and empowers developers to be as productive as possible, contrary to the typical for the financial sector blown-out processes and approval chases. This solution is powered by Google Kubernetes Engine (GKE) and the cloud's niftiest security tools from the Secure Supply Chain toolkit.

BigQuery allows you to generate multimodal embeddings and perform vector searches directly on your data without complex preprocessing steps. Simplify the process of finding relevant data, identifying patterns and trends, and clustering similar objects together. Learn how to generate embeddings using familiar BigQuery SQL syntax with multimodal inputs (text, images, audio). We’ll then review how to use BigQuery’s vector search capabilities to explore data in new and innovative ways, leading to faster decision-making and improved insights.

I have worked with many startups which emerged from university projects. They often run on a single virtual machine, which is what works for research. However, as the system scales and enters production, automation becomes crucial. Google Vertex can radically streamline this process. For example, model training is more efficient with a managed service that automatically scales compute resources based on your training needs, eliminating the cost of idle resources – as happens when you use a Jupyter Notebook or a VM for training. We’ll cover other parts of the ML process with Vertex and other Google Cloud services, including development, hypertuning, deployment for inference, tracking model performance, and defining the entire pipeline. The hard part is deciding when to switch. Learning new services and migrating to them takes precious time that a fast-moving startup cannot afford. But If you stay too long with the simple architecture, you end up building components yourself, and eventually get locked in. So, after reviewing the Google Cloud services you can use, I will explain the key decision points for when to adopt each of the services.

In this session, we will became familiar with Gemma 2 by leveraging Kaggle resources: Models and Notebooks. We will go through few notebooks and explore Gemma 2 capabilities while using prompt engineering, RAG and fine-tuning to prototype applications with this compact and powerful model from Google

This talk explores the paradigm of Agentic AI, highlighting the roles of intelligent agents in transforming modern software architecture. We will discuss the essential design patterns, including tool use, multi-agent collaboration and cross-species agent design, explaining how these frameworks enable AI agents to expand their capabilities, collaborate effectively, and adapt to heterogeneous environments. We will draw analogies between AI agents and microservices architecture, showing how agentic workflows can enhance process optimization, adaptability, and innovation. The talk will emphasize the characteristics of reactive, proactive, and hybrid agents, illustrating their roles in dynamic, real-time decision-making and strategic planning. Additionally, we will assess emerging trends in Generative AI, focusing on advancements in collaborative efficacy, adaptability, and the ethical considerations involved in AI deployment, including building responsible AI systems. This session aims to provide attendees with a comprehensive understanding of how agentic design patterns are influencing the landscape of intelligent systems and their integration within existing software frameworks.

So, you need to have your Java app deployed and available for your users fast, and you would like to do as little managing, wiring, and infrastrastructure work as possible. How can you achieve that? Well, you go serverless with your containers using supersonic Java in the Cloud! Join us exploring the possibilities with MicroProfile and Quarkus on Google Cloud Run. Bleeding edge Java!

Introduction to Security Testing - Areas of security present in any entity, how to perform a security test on various levels AI current state review - Status of the AI technology at the date, models available that involve security Areas where AI helps in Security - Tools, practices, approaches enabled by AI Areas where AI introduces challenges to Security - New attack types, skills, impact that AI generates on security Future ahead of us - What is predicted to come in the near and long term future in regards to cybersecurity

Notifications play a pivotal role for any user-facing application as it gives your users the necessary updates when required. However, if they are not used correctly, they can end up being an annoyance for your users as opposed to something they need. Android has Notification Channels which assist in placing notifications into groups or categories, think of, "Urgent Notifications" vs "Promo Notifications". By placing notifications into channels, important notifications can be received into the right channel and non-trivial notifications can be optionally received and the user can decide to receive them, or not, without compromising receiving important notifications. This can be very beneficial for a delivery app for example - where delivery updates are the important notifications and promo notifications are the trivial, which can be turned off.