Technical Perspectives
Blog & Insights
Articles focused on analytics modernization, data platform reliability, and the practical application of agentic AI in regulated engineering environments.
Research investigating the boundaries between what a system can determine, what it can infer, and what it can independently verify — a proposed new direction plus two prior, independently scoped programs. Full research overview.
Autonomous DataOps Research · September 30, 2026
When Better Evidence Doesn't Produce Better Answers
Evidence Packets Phase I closes: a clean null on representation, and a thin, qualified positive signal on preservation. The full RQ1/RQ2 synthesis, standalone.
DICOM Trust Boundary Research · September 18, 2026
Before It Lands
A demonstrated pre-persistence DICOM policy boundary (fastDICOMgateway), pushed through five escalating boundaries against a real object store — and the negative-control testing that found three bugs in the validation apparatus, not the system under test.
DICOM Trust Boundary Research · August 13, 2026
Don't Land What You Don't Want to Own
Why the safest place to remove PHI from a DICOM object may be pre-persistence, not downstream — validated against ~26,600 real-world CT DICOM objects with fastDICOMattrs.
Autonomous DataOps Research · August 3, 2026
Designing Evidence Packets
How a dbt-specific Evidence Packet became a source-neutral architecture: dbt as the first source adapter, explicit evidence-completeness gaps instead of silent omissions, semantic compatibility with the original packet version, and deterministic packet integrity.
Autonomous DataOps Research · July 27, 2026
Logs Aren't Enough
Why pipeline diagnosis begins with an evidence scavenger hunt—and why structured Evidence Packets should become first-class operational artifacts.
July 2026
Policy-Governed DataOps Agents
Why the layer above orchestration shouldn't be another scheduler or transformation engine, but an operational reasoning layer that separates policy, evidence, judgment, and action.
June 2026
The Deeper Truth: Platform Complexity
A follow-up reflection on when additional data platform complexity is justified, and how to distinguish necessary complexity from premature or accidental complexity.
May 2026
Why I Prefer BigQuery, dbt, and Airflow
A case for letting the cloud-native data plane scale while keeping the control plane understandable.
May 2026
The GCP Swamp
A visual guide to modern GCP-native data platforms as an ecosystem of storage, orchestration, messaging, compute, and analytics.
April 2026
AI-Native SDLC
Designing an SDLC that treats AI agents as bounded contributors within a deterministic orchestration framework.
Jan 2026
Scaling Business Intelligence
Moving from spreadsheet-heavy reporting to a durable, cost-aware data platform without a disruptive rewrite.
Dec 2025
Finding Signal in Events
How observable events propagate into measurable attention patterns, and why disciplined ingestion matters.