Sridhar is a seasoned AI and Cloud Engineering leader with over 14 years of experience driving enterprise-scale technology transformations across the financial services and cloud technology sectors. Currently serving as Senior Manager and Expert AI/ML Platform Architect at Discover Financial Services, he has architected and delivered an enterprise-wide AI/ML and Generative AI Decisioning Platform that reduced model time-to-market by approximately 80 percent and achieved infrastructure cost savings of nearly 65 percent. Holding the highest “Expert” designation on Discover’s internal architecture scale, Sridhar’s platforms directly power real-time Fraud Detection, Anti-Money Laundering, and credit decisioning operations that safeguard millions of customer accounts daily. Sridhar’s expertise is further defined by his pioneering work in Generative AI, where he designed Discover’s “Cleanroom” a secure, compliance-grade environment for fine-tuning and deploying Large Language Models at enterprise scale. Within its first year, the platform he architected supported approximately 50 percent of Discover’s global AI production workloads. Prior to this, his four-year tenure at Microsoft as a Senior Azure Cloud Engineer saw him serve as a trusted technical consultant and crisis resolution owner for the company’s most strategically critical enterprise customers worldwide, earning him the Customer Champion and Top Customer Success Engineer awards. In addition to his engineering achievements, Sridhar is a proven people leader who has directed six cross-functional teams of over 30 engineers across multiple business lines. His unique ability to unite deep technical mastery with commercial acumen designing systems that are simultaneously innovative, secure, and cost-efficient has made him a recognized force in the global AI platform engineering community, including recognition as a Hortonworks Community Guru ranked in the worldwide top 50. With a Master’s Degree in Computer Science from Texas A&M University (GPA 3.83) and an MBA from Bellevue University (GPA 3.90), and credentials including Certified Kubernetes Administrator, AWS Solutions Architect, and HashiCorp Terraform Associate, Sridhar Reddy Bandaru stands out as one of the most accomplished AI platform leaders in the industry a nominee for the Star Award recipient whose work continues to define the standard for secure, scalable, and intelligent enterprise AI.

From Black Boz to Glass Box: Making AI Agents Explainable AI agents don't fail the way traditional software does. Their mistakes unfold across chains of reasoning, tool calls, memory lookups, and model invocations that cross frameworks and services and when something goes wrong, the evidence sits in logs that show what happened but never why. Traditional application performance monitoring explains services, endpoints, and infrastructure; it has no vocabulary for agent behavior. As organizations move agentic AI from prototype into production, that gap becomes a real liability: incidents take too long to resolve, quality regressions slip through unnoticed, and leaders have no reliable way to know whether an agent system is safe to ship.   AgentTrace closes that gap. It is an enterprise observability platform purpose-built for AI agent workflows, turning every agent execution into an explainable workflow graph that updates in real time not a wall of disconnected logs. Rather than guessing at what a system did, teams see the full path an agent took: which tools it called, which models it invoked, which memory it touched, and which policies it triggered, reconstructed across SDKs, OpenTelemetry, MCP, and agent-to-agent protocols into one coherent trace. AgentTrace goes further by showing which execution layers were actually invoked versus merely inferred, giving teams a confidence-graded view of agent behavior even when telemetry is incomplete, thanks to built-in static discovery and ready-made adapters for the frameworks teams already use.   Operationally, this changes how teams work. Root cause analysis shifts from digging through scattered logs to guided, layered drilldown and run replay, with behavior changes linked back to the exact code that caused them. Quality iteration becomes measurable, with every run compared against trusted baselines Golden Traces inside an evaluation workbench that scores runs, clusters failures, and recommends improvements. Releases stop being a leap of faith: a release gate rolls up baseline comparisons, evaluation results, cost, and alerts into one clear pass, warn, or fail decision before code ships. Because AgentTrace is designed with governance in mind tenant-scoped access, sanitized evidence, and privacy-safe exports that never persist raw prompts, completions, or secrets enterprises can adopt it without compromising security or compliance.   Just as application performance monitoring became indispensable infrastructure for the era of microservices, AgentTrace is built to become the essential observability, evaluation, and release-confidence layer for the era of agentic AI. For teams betting their products, and their trust, on autonomous agents, it offers something traditional tooling never could: the ability to see, explain, and stand behind what their agents actually do.