ptokito

AWS (Bedrock, CloudWatch, IAM), Terraform, Python, Kubernetes, OpenTelemetry. Background in observability and enterprise technical customer work.

AI governance and evaluation

A series of labs testing whether standard observability catches AI failures. Most recent one measures how often a Bedrock guardrail blocks questions it was never meant to block.

Conversations about AI evaluation, guardrails and governance. Happy to compare notes with anyone testing AI safety controls or working on observability for AI systems.