Yesu Vara Prasad Kollipara

Assured, Explainable, and Auditable AI for High-Stakes Decision-Making

Abstract:

Artificial intelligence is increasingly being used in high-stakes domains such as healthcare, criminal justice, finance, and public administration, where decisions directly affect human lives and require strong accountability. However, many modern machine learning models operate as black boxes, making it difficult to understand, validate, and trust their outputs. This session explores practical approaches for building trustworthy AI systems that are explainable, reliable, and auditable. It brings together key techniques across interpretability, uncertainty estimation, fairness evaluation, and operational governance. We will compare post-hoc explanation methods, such as feature attribution and counterfactual reasoning, with inherently interpretable models, highlighting when each approach is appropriate. The session also introduces causal reasoning as a way to move beyond correlations toward more meaningful, intervention-oriented insights. To address reliability, we will discuss methods for quantifying uncertainty, including calibrated probability estimates and conformal prediction, which help bound errors in safety-critical settings. We will also examine fairness auditing techniques and the challenges of balancing accuracy with equity across different groups. Finally, the talk covers operational practices required for real-world deployment, including dataset shift detection, continuous monitoring, model versioning, and documentation frameworks such as model cards. We will also reflect on current limitations, including the difficulty of scaling explainability to complex models and aligning machine-generated reasoning with human decision-making. Overall, this session provides a structured, end-to-end view of how to design AI systems that support responsible and verifiable decision-making in environments where failure has significant consequences.

Profile:

Yesu Vara Prasad Kollipara is a highly accomplished Senior Full Stack Lead Developer with over 14 years of advanced software engineering experience, specialising in enterprise-scale, cloud-native, and API-driven applications. His technical expertise spans Java/J2EE, Spring Boot, Microservices, Angular, React, AWS, and modern DevOps tooling, enabling him to lead complex digital transformation initiatives across financial, government, and enterprise sectors. Prasad currently contributes to KeyBank as a Senior Full Stack Lead Developer, where he architects and modernises commercial banking platforms, including Virtual Account Management and Virtual Commercial Cards. His work focuses on secure microservices, API-driven treasury operations, OAuth2/JWT authentication, and seamless integrations with embedded banking partners, driving KeyBank’s modernisation and innovation strategy. Previously, at the Centres for Medicare & Medicaid Services (CMS), he led UI modernisation efforts, Angular migrations (9→11→17), Spring Boot API development, enterprise security enhancements, and Section 508 accessibility compliance. His earlier roles at Virginia State Police, Caterpillar, Freddie Mac, K-12 Inc., and Cisco (via TCS) strengthened his expertise in scalable front-end design, backend engineering, ORM frameworks, cloud readiness, and CI/CD automation. Prasad holds a Master’s in Information Systems, a Master’s in IT Management, and is pursuing a PhD in Information Systems (AI/Cybersecurity) at Dakota State University. Known for innovation, leadership, and for delivering secure, high-performance solutions, Prasad excels at bridging business goals and modern engineering practices to build resilient, customer-centric applications.