From Conversation to Application: Create Your First Cybersecurity Analysis AI Agent with Google ADK + Local LLM

This event will guide participants to hands-on experience with Google’s open-source framework ADK (Agent Development Kit), combined with a large language model running on their own computers. Unlike chatbots that only respond, we will build an AI Agent with just a few dozen lines of Python code that can take action itself. Just say a sentence, and it can classify a large batch of website access logs, identify suspicious attacks, and generate a cybersecurity report with a risk score.The course will focus on three core themes. First is Local-first, where the model runs on your own computer through Ollama, ensuring that sensitive security logs do not need to be sent to the cloud, eliminating API costs, and enabling smooth offline operation—this is an essential requirement when handling real cybersecurity data. Second is Code-first, where we define the Agent through code, making it testable, version-controlled, and suitable for team collaboration. Finally, we have an immediate-use visual interface, adk web, where you can converse with the Agent in the browser right after writing your code, as well as open a tracking history to see step-by-step how it thinks and decides which tools to call.We will personally assemble the model, tools, and commands like building blocks. Through this course, participants will understand the fundamental differences between AI Agents and regular chatbots, grasp why sensitive data applications must choose local large language models, and learn to dismantle real scenarios into a design mindset of tools doing tasks and models making judgments, transforming scattered ideas into a truly operational AI Agent that keeps data in-house and can continuously expand.