About PDRAM

PDRAM is a publication about legal technology, public defense, and artificial intelligence — written from inside the practice, not above it.

The name stands for Public Defender Rage Against the Machine. It's also a reference to RAM — the memory that makes processing possible. PDRAM is specialized memory for public defense: the institutional knowledge, the case heuristics, the evidence patterns that attorneys develop over thousands of cases but that no vendor has built infrastructure to run on.

What This Publication Covers

Analysis — Original long-form writing about legal technology, AI in criminal defense, evidence processing, and the gap between what commercial legal tech builds and what public defenders actually need.

Curated Links — External articles with editorial commentary. Selected daily from investigative journalism, academic research, court filings, and advocacy organizations.

Feed — Aggregated RSS from 27+ sources across legal tech, public defense, AI and law, criminal justice policy, and open-source tools. Updated every 30 minutes.

Community — Discussion threads on any published content. Open to registered readers.

Who Writes This

PDRAM is written by David Karpay, an Assistant Public Defender at the 15th Judicial Circuit Public Defender's Office in Palm Beach County, Florida. He handles criminal defense cases and develops AI tools for evidence processing, local inference, and case management within his office.

Before law, David studied philosophy at Goucher College and completed a back-end web development program at Betamore in Baltimore. The combination of a philosophy degree, a Rails bootcamp, and a decade of criminal defense practice produced whatever this is.

Epistemic Standards

PDRAM operates under an epistemic verification system that governs how claims are evaluated, tracked, and presented. Every published piece distinguishes between verified facts, attributed claims, plausible inferences, and unresolved details. If the reader cannot tell which is which, the piece has failed.

Source reliability is tracked over time. Confidence is expressed in natural language, not numerical scores. Marketing language is flagged and either attributed or reformulated. The full methodology is documented in the publication's verification system specification.

Contact

Email: [email protected]