Webinar 2026 Writing More Code with AI Agents

AI coding agents can produce what feels like an infinite amount of code, extremely quickly. In research, that is only useful if we can trust the result.

This talk presents a practical approach to using AI agents in research-code workflows without assuming their output is reliable. We will follow a working example from raw data through cleaning, validation, analysis, and visualization, identifying where subtle errors can occur and how we might catch them. The goal is not to delegate scientific thinking or judgment to an agent. Instead, we will use agents to cheaply produce sanity checks, diagnostic plots, test cases, and other supporting work that is often expensive in time or expertise to create.

We will then extend this example into reusable patterns for notebooks, shared pipelines, SLURM jobs, and other research-computing workflows. Along the way, we will introduce practical concepts such as context, workflow memory, skills, automated checks, and human-in-the-loop review.

Basic familiarity with large language models is expected. No prior experience with AI coding agents is required.