TECHNICAL BLOG
Build it. Understand it.
Write it down.
Data platforms, AWS cloud, backend engineering and reliable AI agents.
Practical design articles and short notes, separate from my professional case studies.
8 articles to explore.
One search experience. Current access at every boundary.
A shared search architecture for conversations and content: hybrid retrieval, governed ingestion, typed results and permission-aware release.
Build agent context that keeps its evidence.
An incident-assistant design exercise: bounded retrieval, evidence envelopes and checkpoints that preserve uncertainty.
Evaluate agent outcomes, then inspect the failure paths.
Design an escalation-agent evaluation around observable state, scoped tools, calibrated graders and adversarial fixtures.
Define the boundary before adding autonomy.
A design checklist for giving an agent useful tools without making its responsibilities ambiguous.
A data contract belongs at the boundary.
Why a schema is useful, but compatibility and ownership are part of the agreement too.
A data platform starts before the feature store.
Ingestion, domain ownership, contracts and quality form the foundation for analytics, operations and ML.
Design DynamoDB keys from access patterns.
Start with the requests your application must serve, then reason about keys, skew and secondary indexes.
Retry the request without duplicating the effect.
Why a timeout leaves an ambiguous outcome, and how idempotency changes the recovery story.