Sandhya Subramani demonstrates an agent using the open-source Strands framework to create and load tools while running. She shows simple calculator and character-counting examples and explains the role of a system prompt, editor, shell and loader.
She demonstrates creating specialized subagents for a travel-planning taskA multi-agent system contains multiple AI agents that interact, coordinate, divide work, or influence one another while pursuing tasks. and notes that the example lacks access to live information. Her self-repair and broader production applications are presented as possibilities and experience from her demonstrations.
Subramani discusses evaluating outcomes, tool use and inter-agent communicationEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria., then highlights sandboxed executionAn AI agent sandbox is an isolated execution environment that limits which files, processes, networks, credentials, and external systems an agent can access., constrained permissions and observability as safeguards for dynamically generated code.
Watch on YouTube
