Yesha Patel
Enterprise Solution Architect | AI-Driven Commerce & Customer Transformation @ IBM
Tampa, Florida, United States
Actions
Yesha Patel is an Enterprise Solution Architect at IBM specializing in AI-driven commerce, leadership, and business transformation. She leads enterprise-scale digital transformation initiatives focused on modern commerce platforms, cloud-native architectures, customer experience, and intelligent digital ecosystems.
Her expertise spans enterprise architecture, strategic solutioning, API-first ecosystems, and large-scale transformation programs across Retail, Financial Services, Healthcare, and Manufacturing. Beyond client delivery, Yesha actively contributes to the technology community as a speaker, mentor, judge, and IEEE volunteer focused on AI, innovation, leadership, and digital transformation.
Links
Topics
Securing Agentic AI in the Cloud: Identity, Delegated Authority and Zero-Trust Execution
As AI agents begin to access enterprise data, invoke APIs, coordinate workflows, and take actions across cloud platforms, organizations face a new security problem: autonomous software actors operating with delegated authority.
This session presents a practical architecture for securing Agentic AI in cloud environments. It covers non-human identity, least-privilege access, delegated authorization, policy enforcement, tool governance, behavioral observability, human approval, and failure containment.
Attendees will learn how to separate probabilistic AI reasoning from deterministic security controls and how to apply Zero Trust principles when agents interact with APIs, SaaS platforms, enterprise services, and sensitive data. The session also explores practical patterns for detecting abnormal agent behavior, limiting blast radius, maintaining auditability, and governing autonomous actions across distributed cloud environments.
The goal is to provide security architects, cloud engineers, and technology leaders with a vendor-neutral framework for moving Agentic AI from experimentation into secure enterprise operation.
From Legacy Code to AI-Native Software Factories: Modernizing Mission-Critical Systems with Governed
Modernizing large enterprise codebases is no longer simply a matter of rewriting legacy applications in newer languages or moving workloads to the cloud. The harder problem is preserving business behavior, operational resilience, integration contracts, security controls, and decades of embedded domain knowledge while dramatically accelerating transformation.
This session presents a practical architecture for building an AI-driven software modernization factory for large, mission-critical systems.
Drawing on real-world experience modernizing enterprise financial platforms and applied research in AI-assisted software engineering, the session explores how generative and agentic AI can be introduced across the modernization lifecycle—from code understanding and dependency discovery to transformation, behavioral validation, integration testing, and production readiness.
The session will cover practical patterns for:
using AI to analyze and document large legacy codebases;
extracting business logic, dependencies, interfaces, and operational behaviors;
creating semantic representations that preserve system intent across modernization;
combining deterministic engineering controls with LLM-assisted code transformation;
validating behavioral equivalence between legacy and modern implementations;
introducing human review and policy gates into AI-generated software changes;
integrating security, observability, testing, and traceability into the modernization pipeline; and
scaling modernization across multiple applications without creating a new generation of technical debt.
A key focus will be the shift from using AI as a coding assistant to treating AI as part of a governed software factory—where models, agents, engineering tools, testing systems, security controls, and human experts operate within a repeatable modernization workflow.
Attendees will leave with a reference architecture and implementation framework for applying AI to enterprise modernization while preserving the reliability, traceability, and architectural discipline required for critical systems.
AI-Driven Modernization of Mission-Critical Systems: From Legacy to Governed Software Factories
Modernizing mission-critical systems is not simply a code-conversion problem. Enterprises must preserve business behavior, integration contracts, security controls, operational resilience, and decades of embedded domain knowledge while accelerating transformation.
This session presents a practical architecture for building an AI-driven software modernization factory for large and complex enterprise systems.
Drawing on experience with financial-services platforms, distributed architectures, and AI-assisted modernization, the session explores how Generative and Agentic AI can support the full modernization lifecycle: code understanding, dependency discovery, business-rule extraction, transformation, testing, behavioral validation, and production readiness.
Participants will learn practical patterns for:
analyzing and documenting large legacy codebases with AI;
extracting business logic, interfaces, dependencies, and operational behavior;
creating semantic representations that preserve system intent during transformation;
combining deterministic engineering controls with LLM-assisted code generation;
validating behavioral equivalence between legacy and modern implementations;
introducing human review, security, and policy gates into AI-generated changes;
integrating traceability, observability, testing, and quality controls into the modernization pipeline; and
scaling modernization across application portfolios without creating a new generation of technical debt.
A key theme is the transition from using AI as an individual coding assistant to treating AI as part of a governed software factory in which models, agents, engineering tools, testing systems, security controls, and human experts operate within a repeatable lifecycle.
Attendees will leave with a reference architecture and decision framework for applying AI to enterprise modernization while preserving the reliability, traceability, and architectural discipline required for critical systems.
From GenAI Pilots to Production: Securing the Agentic AI Control Plane
Generative AI has moved quickly from experimentation to enterprise adoption, but the next stage is significantly harder: enabling AI agents to reason, invoke APIs, interact with enterprise systems, coordinate workflows, and take action within complex business processes.
Traditional enterprise architectures were not designed for autonomous software actors.
This session presents a practical, vendor-neutral framework for moving from isolated Generative AI pilots toward governed, observable, and production-ready agentic AI.
Drawing on lessons from mission-critical financial-services architecture, large-scale enterprise modernization, and applied work in agentic systems, the session introduces the concept of an Enterprise AI Control Plane: an architectural layer that connects models and agents with the enterprise controls required for secure and reliable operation.
Attendees will explore practical implementation patterns for integrating AI agents with existing APIs, microservices, event streams, data platforms, and legacy systems while preserving the architectural discipline required in regulated and high-availability environments.
The session will examine how organizations can:
Establish agent identity, delegated authority, policy enforcement, and least-privilege access.
Introduce human approval and escalation controls for high-risk or irreversible actions.
Implement observability across prompts, tool calls, workflows, decisions, and autonomous actions.
Evaluate agent behavior, reliability, and operational risk before production deployment.
Maintain semantic consistency when agents interact with fragmented enterprise systems and data.
Design resilient workflows that account for model failures, API failures, downstream outages, and uncertain agent behavior.
Integrate Generative and Agentic AI incrementally without requiring wholesale replacement of existing enterprise platforms.
Create clear governance boundaries between AI reasoning and enterprise execution.
A practical reference architecture will illustrate how models, agents, enterprise services, APIs, identity, policy, observability, security, data, and human governance can operate as a coordinated system.
The session will also address one of the most important architectural shifts created by agentic AI: enterprise APIs are no longer consumed only by applications and people. Increasingly, they will be consumed by autonomous agents acting on behalf of users, systems, and business processes. This changes how organizations must think about authentication, authorization, traceability, lifecycle management, operational resilience, and risk.
Rather than presenting another AI platform or model comparison, this session focuses on the architectural and operational capabilities enterprises need regardless of which models, clouds, or AI platforms they adopt.
Attendees will leave with three practical outcomes:
A reference architecture for connecting AI agents safely to enterprise systems, APIs, data, and workflows.
A governance model for agent identity, authority, observability, human oversight, and operational risk.
An incremental modernization approach for introducing agentic capabilities into existing enterprise environments without disrupting mission-critical systems.
The objective is to help technology leaders move beyond the question of “How do we use Generative AI?” toward the more important enterprise question:
“How do we allow intelligent systems to act within the enterprise while keeping those actions secure, governed, observable, resilient, and accountable?”
The session provides attendees with a practical architecture and decision framework they can apply when moving from AI experimentation toward production-scale Agentic AI.
Your Agent Has a Credit Card Now: Zero-Trust Patterns for Agentic Commerce
*Applicable to everyone who shops online*
AI agents are about to start spending real money on behalf of real users — and most of our security models weren't built for it. When an agent browses a product page, reads a review, and clicks "buy," who is the principal? Who authorized the spend? What stops a prompt-injected review from draining a user's card?
This session walks through zero-trust patterns for agentic commerce: scoped and short-lived credentials, capability-based tool access, spend caps and velocity limits, intent verification, and audit trails that survive an LLM's account of what happened. We'll look at real threat models — confused deputies, prompt injection via product data, runaway autonomous spend — and the emerging standards (OAuth token exchange, Stripe and card-network agent payment specs, MCP authorization) developers can build on today.
ReconGraph: Agentic AI for Real-Time Trade Break Discovery in Enterprise Securities Operations
What happens when an agentic AI system is responsible for resolving trade breaks across 103 million daily transactions — with zero tolerance for error and a T+1 regulatory clock ticking?
This session introduces ReconGraph, an agentic graph-reasoning framework purpose-built for post-trade reconciliation in enterprise securities operations. Unlike traditional rule-based exception management, ReconGraph deploys autonomous AI agents that traverse a live knowledge graph of trades, counterparties, custodians, and settlement states — discovering break patterns, classifying root causes, and recommending resolution pathways with full audit traceability.
Attendees will walk through the architecture of a production-grade agentic system operating under FINRA and SEC compliance constraints, including how agent decision boundaries are enforced, how human-in-the-loop escalation is triggered, and how every agent action is logged to satisfy regulatory audit expectations. This is not a proof-of-concept — it is a practitioner's blueprint from the front lines of T+1 settlement modernization.
Key takeaways: graph-augmented agent design patterns for financial workflows, compliance-aware agent boundaries, and lessons from deploying autonomous AI in zero-downtime fault-tolerant infrastructure.
From Ethics to Control: Building Auditable AI Agents for Regulated Financial Enterprises
Every financial firm is deploying AI agents. Very few can answer the question regulators are starting to ask: "How do you know what your agent did — and why?"
This session presents the Ethics-to-Control Trace Graph (ECTG), a governance architecture designed specifically for agentic AI operating in regulated financial environments — wealth management, capital markets, and enterprise securities operations. ECTG creates a live, executable mapping between AI ethics principles and operational control points, transforming agent governance from policy documents into auditable runtime behavior.
Drawing on research aligned with the emerging IEEE 3410-2025 Model Risk Management standard for GenAI in finance, this talk demonstrates how enterprise architects and AI engineers can instrument their agent pipelines for explainability, control, and regulatory defensibility — without sacrificing the autonomy that makes agents valuable.
From AI Side Project to Production
From AI Side Project to Production: How to Actually Scale GenAI Applications
Everyone has built a ChatGPT wrapper. Few have scaled one.
In session, we’ll break down what separates AI demos from production-ready systems. You’ll learn:
- Why most AI pilots fail to scale
- How to design a scalable GenAI architecture (APIs, orchestration, observability)
- Cost control strategies for LLM-based applications
- Governance, security, and compliance considerations
- Measuring ROI beyond "it sounds cool"
Using real-world commerce and enterprise use cases, we’ll walk through how to move from hackathon-grade prototypes to durable, enterprise-grade AI platforms.
Perfect for developers building AI products, founders integrating LLMs, and engineers modernizing legacy systems.
Building for the Agent Shopper: What Changes in Your Stack When AI Agents Buy
*Applicable to everyone who shops online*
Commerce platforms were built for humans — browsing products, comparing options, and completing purchases manually. But as AI agents evolve into autonomous decision-makers, enterprises must rethink how digital commerce systems are designed and governed.
This session explores the rise of the “agent shopper” — AI systems capable of discovering products, evaluating options, and making purchasing decisions on behalf of users or businesses. Attendees will learn how enterprise architectures must evolve to support agent-driven commerce through intelligent APIs, event-driven workflows, secure integrations, and real-time decisioning.
The talk also covers governance, trust, fraud prevention, and scalable architecture patterns needed to build secure, AI-ready commerce ecosystems for the next generation of digital interactions.
Architecting AI-Ready, Event-Driven eCommerce with Secure Microservices at Scale
Modern digital commerce is no longer just transactional — it is intelligent, distributed, and real-time. Retail organizations are re-architecting legacy monoliths into API-first, event-driven ecosystems capable of supporting AI personalization, dynamic pricing, and omnichannel fulfillment.
In this session, we will explore best practices for designing cloud-native eCommerce platforms using domain-driven microservices, asynchronous messaging, and zero-trust security principles. Drawing from enterprise retail modernization programs, this talk will cover:
Decomposing monolithic commerce platforms into bounded contexts
Designing scalable REST, GraphQL, and event-stream APIs
Implementing resilient service-to-service communication
Embedding AI capabilities into commerce workflows without compromising latency
Securing APIs with OAuth2, tokenization, and runtime threat detection
Observability strategies for distributed commerce systems
Attendees will leave with architectural blueprints and practical patterns to build AI-ready, secure, and highly scalable commerce platforms that operate reliably during peak demand events.
Why this works:
Strong alignment with microservices + AI + cloud-native modernization — high relevance to 2026 audience.
Yesha Patel
Enterprise Solution Architect | AI-Driven Commerce & Customer Transformation @ IBM
Tampa, Florida, United States
Links
Actions
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top