Blog

Technical notes on infrastructure, automation, and building tools that stay useful even when the AI layer isn't available.

  • Rolling out distributed tracing without breaking what already worked

    Onboarding a Java microservices platform onto distributed tracing without asking any team to touch their code, and without disturbing the logging and metrics pipelines already running in production.

  • Filing my first patent: what the process actually taught me

    The invention itself is still confidential while the application is pending, so this isn't a technical write-up. It's what I learned going through prior-art research and defending the idea to a patent attorney for the first time.

  • Cloud cost optimization: three initiatives that actually moved the needle

    Reserved-capacity planning for DynamoDB, traffic-aware rightsizing for Aerospike, and a unified multi-cloud cost dashboard -- three different approaches to the same underlying problem: spend that had drifted away from actual usage.

  • One chat interface instead of five tools: building Lens

    Investigating an issue across metrics, logs, and a database usually means several open tools and a manual correlation step. Lens replaces that with one natural-language chat interface — here's the architecture behind it.

  • Filling a gap in the Terraform MongoDB Atlas provider

    Why I published terraform-mongodbatlas-indexes: declarative, YAML-driven MongoDB Atlas index management, backed by unit tests, a live-database integration test, and CI on every push.

  • Finding the process behind a noisy VM: AI-based root cause analysis across a fleet

    Cortex VM Analyzer compares a suspect host's VM- and process-level metrics against a healthy peer, and uses AI-based analysis to name the specific process behind a resource bottleneck or memory leak -- before anyone opens application logs.