Ravikumar Buragapu
I lead AI engineering and infrastructure for Oracle Health, where I work on bringing agentic AI into large, regulated platforms. Before that I built developer platforms and AI-native engineering programs at Adobe and Wells Fargo.
I write about what it takes to run AI agents safely in production: identity, evaluation, observability, and the controls around them. My latest piece isThe Hard Part of Agents Isn't the Model.
Last time I said autonomy should be earned. Here is how I’d actually measure it: replayable traces, component-level attribution, planted failures for reviewers, and error budgets that raise or lower an agent’s tier.The Hard Part of Agents Isn't the Model.