HIDA Lecture: Agentic AI - Safe Autonomous Agents
Speaker: Jan Ebert, NVIDIA
Date: 16.06.2026, 11 am
Title: Agentic AI: Enabling Safe and Productive Autonomous AI Systems with Coders and Claws
Abstract
Toward the end of 2025, autonomous AI agents made significant leaps in capability. Smarter coding agents, a growing ecosystem of integrations exposed as skills and Model Context Protocol (MCP) servers, and projects like OpenClaw have demonstrated how these components can be combined into self-evolving AI systems. This lecture explores what these developments mean in practice, how to leverage existing tools to build new capabilities, and how they can be applied to automate workflows and boost productivity.
A key focus is security. Autonomous systems introduce new risks, and balancing safety with sufficient autonomy is a delicate challenge. We provide a practical overview of core principles for working safely with AI, using existing safeguards where possible, and discuss responsibilities as well as when human oversight (“human in the loop”) remains essential.
While the lecture emphasizes software and research perspectives, it also touches on organizational use cases where AI can provide support. The insights are designed to be transferable across domains.
Jan Ebert
Jan Ebert is a Senior Software Engineer at NVIDIA, developing its open source large language model (LLM) frameworks under the Megatron-LM and NeMo umbrellas. He previously co-founded LAION e. V. and collaborated on the release of free, open models in projects such as OpenGPT-X and TrustLLM while working at Jülich Supercomputing Centre. His main areas of interest are secure automation through AI, parallelization, and optimizing resource efficiency for AI training and inference.
