Reliable AI delivery isn't either/or—it's both/and. Test conventionally for functionality. Evaluate probabilistically for quality. Deploy with dual-discipline confidence.
Build a Slack bot using AWS Bedrock and MCP to answer GitHub questions. Learn setup, architecture, and how to extend it with new tools and data sources.
Building a Slack bot with traditional APIs led to 400 lines of code. Using MCP and AWS Bedrock reduced complexity, enabling scalable, tool-driven automation.
Setting up a data catalog isn’t just a tool problem. My work with Azure Purview and Collibra showed success depends on governance, metadata, and adoption.
AI generates code faster than tests can cover. Coverage stays green while gaps grow. Treat AI code as untested by default and scale testing to match generation speed.
Unbounded retries and autoscaling can turn minor latency into cascading outages. API reliability must be bounded and load-aware to prevent retry storms.
SAP cloud TCO is driven more by landscape sprawl than by EC2 costs; optimize environments and use Terraform, S3, and EFS lifecycle policies to reduce costs.
Retesting isn’t a checkbox — it’s discipline: reproduce, verify fixes, test edges, run regression, validate in staging, document, automate, and never skip it.
Learn how agentic testing reshapes QA by adding governance, traceability, and accountability to AI-driven workflows, ensuring speed doesn’t compromise quality.
Multi-cloud sounds strategic, but usually happens by accident. Networking, IAM, and observability all break at boundaries. Only attempt it if you have no choice.
Three protocols are shaping how AI agents interact with tools, other agents, and users. Here's what each one does, how they fit together, and when to reach for which.
Strong QA is not checklist work. It combines investigation, analytical thinking, and technical communication to find failure paths early and improve the system over time.