On October 23, CADS hosted Agentic AI in Action: From Idea to Production, a hands-on workshop led by Naveen Ramakrishna, Senior Principal Software Engineer at Dell Technologies.
The session was built around one idea: autonomy is a budget. Use an AI model where the work needs language or judgment, use code where there is a correct answer, and require a human wherever something becomes official.
Working in a shared notebook, participants engineered a “DataCard agent” that documents research datasets. Over five labs they:
- decided when a problem needs an agent, a workflow or a simple script, and watched a small agent choose its own steps;
- compared AI models on quality, speed and cost, and saw that output can pass every format check and still be wrong;
- built rule-based evaluation with a capped AI judge and a bounded fix loop;
- ran a batch under real rate limits, timeouts and failures, and diagnosed them from a trace;
- carried out a prompt-injection attack hidden inside the data, then defended against it with layered guardrails and a human approval step.
The slides and notebook are available below.