Nagurva Pamuluri

Thrilled to announce our 3rd Virtusa | AWS re:Connect '26 event! Building scalable, governed, and intelligent AI across the enterprise is more vital than ever, and I'm looking forward to hosting our AWS leaders, partners, and clients in Hyderabad later this month. #AWS #Virtusa #AgenticAI #EnterpriseAI #VirtusaAWS #AWSreconnect2026

16

What if your AI agent could rewrite its own brain — and teach itself new skills — after every run? Most agent frameworks are static. You craft a prompt, wire up tools, and hope for the best. But real intelligence adapts. I've been building a self-evolving deep agent that runs a closed-loop improvement cycle after every execution. Three independent graders score each trajectory on task completion, efficiency, and output quality. From high-scoring runs, the system automatically extracts reusable skills — stored as versioned artifacts the agent can invoke in future tasks. Failures trigger automatic prompt rewriting based on analyzed patterns. The compounding effect is powerful. Each cycle produces an improved prompt, new learned skills, and enriched episodic and semantic memories. The agent doesn't just get corrected — it acquires new capabilities from its own experience. Score improvements from 0.65 to 0.78+ in just 2-3 cycles. Shouldn't self-improvement and skill acquisition be first-class capabilities in production agentic systems? https://lnkd.in/eVBdF7bR #AgenticAI #LangGraph #DeepAgents #OpenSource

59

1 Comment

This blog post that covers how cloud computing is evolving into an agentic ecosystem. It discusses how AI agents are autonomously orchestrating migration, modernization, operations, and optimization. It highlights how this shift delivers self-healing capabilities, continuous optimization, faster innovation, lower costs, and reduced operational complexity.

39

War and geopolitical conflict disrupt lives in ways we often don’t see. My thoughts are with everyone impacted by the ongoing situation in the Middle East. 🙏 But crises also reveal how organizations truly treat their customers. Right now, my spouse is stranded in India because of a disrupted flight with Qatar Airways , booked through Expedia . The disruption due to the war is understandable. Safety always comes first. What is not understandable is the complete breakdown in customer support that followed. Here’s the reality customers are facing: • No proactive rebooking on partner airlines from Qatar Airways • No confirmed refund yet • Redirected to the booking agent because the ticket was issued through Expedia And contacting the agent? • 1.5 hours wait to speak with a human • ~20 minutes wait just to reach a virtual chat agent • Two separate chats with different agents — still no refund confirmation Meanwhile, a traveler remains stranded thousands of miles away from home. This raises a bigger question for the travel industry. Airlines and OTAs constantly talk about AI, automation, and digital transformation. But when real disruption happens — war, weather, operational shutdowns — why does customer experience collapse? Disruptions are inevitable. How companies respond during disruption defines the brand. Travelers need: • Proactive rebooking options • Faster refunds • Real-time disruption management • Human + AI support that actually works Moments like this are when customer trust is either built… or lost. Hoping Qatar Airways and Expedia in quickly to resolve this situation and strengthen how disruption support works for customers. #CustomerExperience #TravelDisruption #Airlines #AI #DigitalTransformation #TravelIndustry #qatarairways #expedia Ariane Gorin Hamad Al-Khater Jochen Koedijk Shilpa Ranganathan

43

4 Comments

Giving an AI agent unlimited access to your enterprise systems is like handing a new hire the master key on day one. In production, agentic AI must be architected with controlled tool access, least privilege, and explicit guardrails. With Amazon Bedrock Agents, you can let an agent retrieve data and call approved APIs, but gate high-impact actions behind authorization or human approval. A simple chatbot only answers questions; an agent can change state—update records, trigger workflows, modify configurations. That usefulness is also the risk. The architectural challenge is to make the agent powerful enough to be valuable, but constrained enough that a bad decision or prompt injection can't cause real damage. Scope its tools, enforce policies, log every action, and require approval for anything that touches money, data, or infrastructure. If your AI agent could take one irreversible action today, what would you want to be sure it cannot do without human sign-off? #AgenticAI #AWS #Bedrock #EnterpriseArchitecture #AIGovernance #CloudSecurity

4

We're excited to announce our partnership with Virtusa Corporation. Virtusa will deploy PlayerZero's AI production engineers across enterprise clients and private equity portfolio companies worldwide, bringing autonomous agents capable of operating production software in support, QA, and modernization. Engineering teams today are stuck in reactive mode, firefighting tickets and incidents while institutional knowledge stays trapped in the heads of a few senior engineers. Our partnership helps change that. PlayerZero brings AI production engineers powered by the Production World Model, a context graph that captures how people, process, and technology work together. Virtusa brings global delivery scale, governance, and deep vertical expertise to help enterprises bootstrap and scale that model. This is about making institutional knowledge durable and scalable. Read more about what we're building together: https://lnkd.in/gwrCBTNk

51

3 Comments

Every hour we take out of an authorization turnaround is an hour a member gets back on the way to their care. That is the framing I hold onto in this work. Our teams have spent years getting faster at this, and the progress is real. Speed to care is the metric that matters most to the people we serve, and it is the reason I pay close attention to any technology that can move it meaningfully rather than incrementally. Agentic AI looks like one of those technologies, and what makes it genuinely usable is the discipline around it. If an AI system participates in a coverage determination, we should be able to explain that determination clearly to a regulator, to the treating provider and to the member. Explainability is not a constraint on this work. It is what earns it the trust to scale. I am joining a Healthcare Dive panel on October 22 with colleagues from NASCO, Healthfirst and Pega to work through both halves of that. I would value your perspective in the Q&A. https://lnkd.in/ep3Ns3TY #HealthcareAI #AIGovernance #PriorAuthorization #MemberExperience

52

Legacy modernization is shifting from automated refactoring to AI-led re-engineering. As costs and timelines converge, line-by-line code conversion is proving insufficient to deliver scalability, resilience, and long-term value. AI-led re-engineering enables deeper architectural change by combining domain understanding, cloud-native design, and modern engineering patterns.

22

I had an amazing time at the AI Summit in Boston last week. The sessions were incredibly insightful, and I had the opportunity to network with peers who are leading the next wave of AI-driven transformation. I came away with numerous takeaways on applying agentic AI for real-world enterprise impact, along with practical perspectives from industry leaders. 1.     AI in the real World:  It was impressive to see how Waymo uses AI to train and simulate their models, for various  teacher and student models. @Smita Shyam covered how the building synthetic data systems, rapid training achieved using simulation (matrix style) and how edge cases are simulated. 2.    Secure AI by Design: It was eye opening to see various levels of security vulnerabilities brought in by the AI systems and how an unified platform starting from and browser may be required to protect users and applications - ensuring every agent and models operates with integrity, transparency, and accountability.  Similarly, threats posed by autonomous agents (including MCP) to enterprise security were discussed. 3.    Impressive demos and products covering anything from Collaborative and cooperative agents and user-friendly AI workflow platforms provided insights into DIY mode and No-code mode of agentic deployments Appreciated the insightful discussions with peers and panelists over Human oversight in AI, safety, explainability, alignment, and trust; Balancing innovation, risk and ethics; Delivering real business impact and measurable ROI. It was great to connect with Andrew Amann, Smitha Shyam, Rajkumar Joghee Bhojan, Trang Nguyen , Nathan Bittinger, Anil Sharma, Prashant Desai, Nirupam Sarkar, Ph.D., Kevin Sieck, Josep Carner Marsal, Shreesha Jagadeesh, Brendan Murphy, Maryam Najafian, PhD, Manajit Barman, Jon Boyer Thanks for the opportunity #aiai team. #AI #AgenticAI #ChiefAIOfficerSummit

35

1 Comment