Whitney Lee is a Senior Technical Advocate at Datadog who teaches Kubernetes and developer platforms through live learning, visual explanations, and working software. Her projects range from Enlightning, where she drew cloud-native concepts as guests taught them, to cluster-whisperer, an AI agent that investigates Kubernetes applications and limits deployment to an approved platform catalog.
From photography to cloud engineering
Whitney spent more than a decade running a wedding-photography business before leaving it to tour with her brother Jordan’s band, Mutual Benefit, in 2014. After returning to restaurant work, she tried programming at her son’s encouragement. She wrote her first code in January 2019, completed a coding bootcamp that October, and joined IBM as a cloud developer in November. Her career-change account describes a concentrated year of learning full-stack web development followed by work that introduced her to cloud infrastructure.
At IBM, Whitney built proofs of concept for prospective customers. Between projects, she began making lightboard videos explaining technologies such as Kafka and RabbitMQ. She joined VMware’s Tanzu team in June 2021 and created Enlightning, a live show in which guests taught her cloud-native concepts while she drew them on a transparent board. Her photography experience carried into the production, from lighting the board to composing episode graphics.
Whitney rejects the assumption that artistic and technical ability belong to different kinds of people. Photography and music require technical judgment; software gives her room for visual storytelling, humor, and creative problem-solving.
She later joined Datadog to focus on Kubernetes. Her work there includes hosting Datadog Illuminated and Software Defined Interviews, alongside her independent cloud-native education.
Teaching through questions, designing for developers
Whitney makes the learner’s questions part of her teaching. In Enlightning, she invited specialists to teach her unfamiliar subjects live, leaving room for corrections and incomplete understanding. Her approach to technical education includes spelling out acronyms, deciding what knowledge an explanation assumes, and spending extra time on the steps that confused her when she first learned them. She favors creating something she wants to understand: even an unwatched resource can repay the effort through learning.
Her platform-engineering argument applies similar attention to what developers need to know. A database or infrastructure specialist can encode configuration and security decisions into a service; the platform team supplies an interface that application developers can use independently, with visibility into its behavior. The starting point is the delays and repeated work developers face, then the service that would remove them.
Whitney also challenges compulsory platform adoption. If developers avoid a platform, she treats that as feedback about its documentation, awareness, or usefulness. Adoption becomes a test of whether the platform helps its users rather than merely satisfying the team that built it.
AI interfaces to infrastructure
Whitney’s cluster-whisperer project explores how a natural-language interface can help developers investigate and use Kubernetes. Its mechanisms address distinct parts of that task:
Investigation and discovery: Tools inspect resources and logs, while semantic search helps find capabilities available in the cluster. Developers can ask both why an application is broken and what services they could use.
Restricted deployment: The agent can deploy only resource types in an approved platform catalog. Code enforces that restriction, rather than leaving the decision to the model.
Execution tracing: OpenTelemetry records the agent’s execution so its actions can be inspected.
For an interactive platform demonstration developed with Viktor Farcic, Whitney built the demo to accommodate audience-selected tool combinations. The agent investigated a broken application, searched cluster resources, and deployed on a developer’s behalf. The experiment asked whether adding AI actually improved the developer’s experience.
In Build a Platform and Watch It Burn, Whitney and Michael Forrester explored what happens when an agent receives excessive access. Their deliberately permissive burrito-ordering assistant could deploy an external container image and expose a secret recipe from its Kubernetes environment. They repeated attacks with infrastructure controls and AI input and output filtering, showing both successful restrictions and protections that failed to stop access consistently.
The joint workshop made the limits of each approach visible: model refusals varied between attempts, infrastructure rules constrained deployment, and filtering could block requests or redact sensitive responses. OpenTelemetry traces in Datadog exposed prompts, tool calls, and results, including an attempt to access internal configuration through instructions embedded in a weather tool’s description. These examples extend Whitney’s work on developer interfaces to a harder question: what an AI interface can actually do with the access it receives, and how engineers can inspect and constrain that behavior.
Michael Forrester and Whitney Lee turn a burrito assistant into a Kubernetes attack interface, then test where infrastructure policy, input and output filtering, and tool controls actually intervene.
A persuasive business story changed BurritoBot's willingness to deploy an image; Kubernetes admission policy supplied a separate decision about whether deployment was allowed.
Runtime detection followed by a response differs from preventing an operation. The workshop's intended filesystem block failed, even though the distinction remains useful.
Output filtering can redact an answer after generation; input filtering can reject a request before main-model processing. Their costs and protection points differ.
▶ Watch ↗