Turning Generative Models into Reliable System Components
Generative models are increasingly embedded in production systems, yet their outputs remain probabilistic and difficult to trust. This talk presents a systems approach to making generative AI reliable, testable, and governable. Drawing from real world implementations, it outlines how to treat model outputs as components within a controlled execution layer rather than final responses. The session introduces patterns such as deterministic validation pipelines, layered output checks for safety and accuracy, and arbitration mechanisms that enforce fixed decision rules. It also covers how to design for replayability, auditability, and consistent behavior under the same inputs. Attendees will learn how to integrate these patterns into existing services without slowing down development velocity. The focus is on practical architecture and measurable impact, including reducing production risk and improving confidence in AI driven decisions. This approach reframes generative models from unpredictable tools into dependable building blocks for critical systems.
About the speaker
Nirmal Jingar
AI & Technology Strategy Director at Humanity + AI
Nirmal Jingar leads the organization’s strategy and public leadership in emerging technologies including AI, AGI, and advanced computing systems. He defines frameworks and standards for responsible adoption and real-world implementation across industry and research communities.
In his role on the board, Nirmal oversees the design of high-impact programs and learning initiatives that shape how technology leaders understand and apply emerging technologies. He advises on technical direction, evaluates external work and partnerships, and contributes to broader conversations on technology policy and governance.
Nirmal is the author of Large Language Models for Intelligent Supply Chain Systems and Artificial Intelligence for Supply Chain Optimization and Decision-Making. He has published multiple peer-reviewed research papers on enterprise AI, multi-agent systems, AI governance, and intelligent decision architectures through IEEE and other international conferences.
A frequent keynote speaker, Nirmal has presented at leading technology and AI conferences including AI4, Home Delivery World, Reuters Events Supply Chain USA, and other global industry forums. He also serves on multiple technology and nonprofit advisory boards and a member of the Forbes Technology Council.
His work bridges research and enterprise practice, helping organizations build AI systems that are reliable, governed, and capable of making trusted decisions at scale.