Your AI Agent Is Probably Wrong — How to Build Ones That Aren’t

We are excited to welcome Sagar Jadhav to the USF Data Science Speaker Series.

About the Speaker
Sagar is a Principal AI Architect at Rubrik, where he builds AI agent frameworks, conversational intelligence platforms, and enterprise governance systems. With more than 20 years of experience, he has designed production-grade AI systems across retail, healthcare, and enterprise software, including platforms processing billions of transactions each month. He is also an active mentor to early-career engineers focused on building trustworthy AI at scale.

Talk: Your AI Agent Is Probably Wrong — How to Build Ones That Aren’t
Everyone is building AI agents. Few are rigorously testing them.

Production AI agents often operate at 70 to 90 percent accuracy. That may sound acceptable, but the remaining gap can mean missed security alerts, incorrect recommendations, or misleading customer interactions. Many teams rely on informal testing instead of structured evaluation and therefore do not truly know how reliable their systems are.
In this session, Sagar shares a practical framework for building AI agents that businesses can trust. He will introduce structured evaluation strategies used at Rubrik, explain when an agent is actually the right solution, and show how production systems combine deterministic guardrails with language model reasoning. The session includes a live walkthrough of a Service Desk Triage Agent and a look at a real evaluation pipeline.
You will leave with a practical playbook for evaluating, testing, and improving AI agents at production scale.

We hope you can join us.

#DataScience #ArtificialIntelligence #AIEngineering #AIAgents #LLM #MachineLearning #USFCA #USFMSDSAI #DataInstitute #TechTalk #AIInProduction