GIDS news media, articles, insights and virtual events educate and illuminate its audiences so they can be fully prepared to deal with the new realities at work and in their professions.
Get the latest event updates, and insights from today's leading voices.
< keynote />
Thu, April 23ArchitectureOpsTech DeepTech
A recent IBM–Morning Consult survey shows that 99% of developers building AI applications are already exploring or building AI agents. Yet, despite this momentum, most organizations still hesitate to deploy agents into production. The core barrier is a trust deficit—driven by concerns around robustness, reliability, safety, and policy alignment. In this keynote, we’ll unpack the hidden pitfalls that prevent AI agents from crossing the chasm from development to real-world use. From adversarial robustness testing—including susceptibility to prompt injection and other attacks—to “natural robustness” failures where small input variations trigger wildly inconsistent outputs, we will explore why traditional evaluation often breaks down. We will also examine how continuous guardrails and runtime policy checks can serve as the backbone for trusted deployment. Attendees will walk away with practical best practices, architectural patterns, and evaluation strategies that enable AI agents to earn the trust of risk, compliance, and engineering teams—ultimately accelerating their path to production.
What You Will Learn
Who Should Attend
< speaker_info />
Dr. Manish Bhide is the Distinguished Engineer and CTO for watsonx.governance -- a product that helps build trust and transparency in traditional ML, Generative AI as well as Agentic AI. Manish has a passion for innovation and is credited with conceptualising several technological innovations which have made an impact on IBM’s products and offerings. Manish is a prolific inventor and has filed more than 100 patents and has published more than 25 research papers in IEEE/ACM conferences. He has a PhD in Computer Science where his PhD thesis dealt with applying Machine Learning to solve problems of streaming data.
