Jyostna Seelam

The Trust Gap: Engineering Autonomous AI Systems for Real-World Reliability

Abstract:

As autonomous AI agents move from experimentation into production, they are increasingly tasked with orchestrating workflows, making decisions, and executing system-level actions. Yet greater autonomy introduces a critical engineering challenge: traditional reliability practices were designed primarily for deterministic software, not probabilistic and adaptive systems that can behave unpredictably or fail in ways that are difficult to diagnose.

This keynote examines the trust gap in autonomous AI through a pragmatic, full-stack reliability framework grounded in enterprise infrastructure, DevSecOps, and modern SRE. It explores emerging agentic failure modes - from non-deterministic execution loops to guardrail failures, and introduces a practical three-tier framework:

Design-Time Trust: Guardrails, policy enforcement, evaluation, and deterministic boundaries. Runtime Trust: AI-native observability, confidence-aware execution, and adaptive human-in-the-loop controls.

Post-Action Accountability: Auditability, state recovery, traceability, and automated post-incident analysis.

Attendees will gain a practical blueprint for moving beyond model-centric evaluation toward system-level verification and operational trust. The keynote demonstrates how trust can be engineered as a measurable, observable, and continuously validated capability across complex autonomous AI systems.

Profile:

Jyostna Seelam is an IEEE Senior Member, SCRS Fellow, and Senior Manager & Engineering Lead with 19+ years of enterprise engineering experience spanning full-stack development, software engineering, AI, cloud infrastructure, SRE, DevSecOps, observability, and platform engineering. She brings an end-to-end perspective on building and operating modern technology platforms.

She has architected platforms supporting 15,000+ engineers, delivered significant improvements in performance, resiliency, and infrastructure efficiency, and led automated remediation initiatives addressing 2+ million security vulnerabilities.

Jyostna is the Founder & Chair of the ACM Richmond Professional Chapter and Co-Founder & Secretary of the IEEE TEMS Richmond Chapter. She has conducted 50+ international peer reviews so far , contributed to online platforms like DevOps.com, DZone, and HackerNoon, and authored a Springer Nature chapter on Trustworthy AI Systems.

She has been an invited speaker at JFrog's swampUP conference in San Jose, participated as a leading panelist alongside senior technology executives and CEOs, and is recognized by ADPList among its Top 10 SRE mentors, reflecting her commitment to mentoring and developing the next generation of technology professionals.

Her current focus is Agentic AI, autonomous engineering, AI-native observability, trustworthy AI, and the evolution of software engineering in the AI era.